![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to pgpdoc2.txt CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [PGP] / pgp / doc|
Return to pgpdoc2.txt CVS log | Up to [PGP] / pgp / doc |
1.1 root 1: Phil's Pretty Good Software
2: Presents
3:
4: ===
5: PGP
6: ===
7:
8: Pretty Good Privacy
9: Public Key Encryption for the Masses
10:
11:
12: -------------------------
13: PGP User's Guide
14: Volume II: Special Topics
15: -------------------------
16: by Philip Zimmermann
17: Revised 6 Mar 93
18:
19:
20: PGP Version 2.2 - 6 Mar 93
21: Software by
22: Philip Zimmermann
23: with
24: Branko Lankester, Hal Finney, and Peter Gutmann
25:
26:
27:
28:
29: Synopsis: PGP uses public-key encryption to protect E-mail and data
30: files. Communicate securely with people you've never met, with no
31: secure channels needed for prior exchange of keys. PGP is well
32: featured and fast, with sophisticated key management, digital
33: signatures, data compression, and good ergonomic design.
34:
35:
36: Software and documentation (c) Copyright 1990-1993 Philip Zimmermann.
37: For information on PGP licensing, distribution, copyrights, patents,
38: trademarks, liability limitations, and export controls, see the
39: "Legal Issues" section.
40:
41:
42: Contents
43: ========
44:
45: Quick Overview
46: Special Topics
47: Selecting Keys via Key ID
48: Separating Signatures from Messages
49: Decrypting the Message and Leaving the Signature on it
50: Sending ASCII Text Files Across Different Machine Environments
51: Leaving No Traces of Plaintext on the Disk
52: Displaying Decrypted Plaintext on Your Screen
53: Making a Message For Her Eyes Only
54: Preserving the Original Plaintext Filename
55: Editing Your User ID or Pass Phrase
56: Editing the Trust Parameters for a Public Key
57: Checking If Everything is OK on Your Public Key Ring
58: Verifying a Public Key Over the Phone
59: Using PGP as a Unix-style Filter
60: Suppressing Unneccessary Questions: BATCHMODE
61: Force "Yes" Answer to Confirmation Questions: FORCE
62: PGP Returns Exit Status to the Shell
63: Environmental Variable for Pass Phrase
64: Setting Configuration Parameters: CONFIG.TXT
65: TMP - Directory Pathname for Temporary Files
66: LANGUAGE - Foreign Language Selector
67: MYNAME - Default User ID for Making Signatures
68: TEXTMODE - Assuming Plaintext is a Text File
69: CHARSET - Specifies Local Character Set for Text Files
70: ARMOR - Enable ASCII Armor Output
71: ARMORLINES - Size of ASCII Armor Multipart Files
72: KEEPBINARY - Keep Binary Ciphertext Files After Decrypting
73: COMPRESS - Enable Compression
74: COMPLETES_NEEDED - Number of Completely Trusted Introducers Needed
75: MARGINALS_NEEDED - Number of Marginally Trusted Introducers Needed
76: CERT_DEPTH - How Deep May Introducers Be Nested
77: BAKRING - Filename for Backup Secret Keyring
78: PAGER - Selects Shell Command to Display Plaintext Output
79: SHOWPASS - Echo Pass Phrase to User
80: TZFIX - Timezone Adjustment
81: CLEARSIG - Enable Signed Messages to be Encapsulated as Clear Text
82: VERBOSE - Quiet, Normal, or Verbose Messages
83: INTERACTIVE - Ask for Confirmation for Key Adds
84: Protecting Against Bogus Timestamps
85: A Peek Under the Hood
86: Random Numbers
87: PGP's Conventional Encryption Algorithm
88: Data Compression
89: Message Digests and Digital Signatures
90: Compatibility with Previous Versions of PGP
91: Vulnerabilities
92: Compromised Pass Phrase and Secret Key
93: Public Key Tampering
94: "Not Quite Deleted" Files
95: Viruses and Trojan Horses
96: Physical Security Breach
97: Tempest Attacks
98: Exposure on Multi-user Systems
99: Traffic Analysis
100: Cryptanalysis
101: Legal Issues
102: Trademarks, Copyrights, and Warranties
103: Patent Rights on the Algorithms
104: Licensing and Distribution
105: Export Controls
106: Computer-Related Political Groups
107: Recommended Readings
108: To Contact the Author
109: Appendix A: Where to Get PGP
110:
111:
112: Quick Overview
113: =============
114:
115: Pretty Good(tm) Privacy (PGP), from Phil's Pretty Good Software, is a
116: high security cryptographic software application for MSDOS, Unix,
117: VAX/VMS, and other computers. PGP combines the convenience of the
118: Rivest-Shamir-Adleman (RSA) public key cryptosystem with the speed of
119: conventional cryptography, message digests for digital signatures,
120: data compression before encryption, good ergonomic design, and
121: sophisticated key management.
122:
123: This volume II of the PGP User's Guide covers advanced topics about
124: PGP that were not covered in the "PGP User's Guide, Volume I:
125: Essential Topics". You should first read the Essential Topics
126: volume, or this manual won't make much sense to you. Reading this
127: Special Topics volume is optional.
128:
129:
130:
131: Special Topics
132: ===============
133:
134:
135: Selecting Keys via Key ID
136: -------------------------
137:
138: In all commands that let the user type a user ID or fragment of a
139: user ID to select a key, the hexadecimal key ID may be used instead.
140: Just use the key ID, with a prefix of "0x", in place of the user ID.
141: For example:
142:
143: pgp -kv 0x67F7
144:
145: This would display all keys that had 67F7 as part of their key IDs.
146:
147: This feature is particularly useful if you have two different keys
148: from the same person, with the same user ID. You can unambiguously
149: pick which key you want by specifying the key ID.
150:
151:
152: Separating Signatures from Messages
153: -----------------------------------
154:
155: Normally, signature certificates are physically attached to the text
156: they sign. This makes it convenient in simple cases to check
157: signatures. It is desirable in some circumstances to have signature
158: certificates stored separately from the messages they sign. It is
159: possible to generate signature certificates that are detached from
160: the text they sign. To do this, combine the 'b' (break) option with
161: the 's' (sign) option. For example:
162:
163: pgp -sb letter.txt
164:
165: This example produces an isolated signature certificate in a file
166: called "letter.sig". The contents of letter.txt are not appended to
167: the signature certificate.
168:
169: After creating the signature certificate file (letter.sig in the
170: above example), send it along with the original text file to the
171: recipient. The recipient must have both files to check the signature
172: integrity. When the recipient attempts to process the signature
173: file, PGP notices that there is no text in the same file with the
174: signature and prompts the user for the filename of the text. Only
175: then can PGP properly check the signature integrity. If the
176: recipient knows in advance that the signature is detached from the
177: text file, she can specify both filenames on the command line:
178:
179: pgp letter.sig letter.txt
180: or: pgp letter letter.txt
181:
182: PGP will not have to prompt for the text file name in this case.
183:
184: A detached signature certificate is useful if you want to keep the
185: signature certificate in a separate certificate log. A detached
186: signature of an executable program is also useful for detecting a
187: subsequent virus infection. It is also useful if more than one party
188: must sign a document such as a legal contract, without nesting
189: signatures. Each person's signature is independent.
190:
191: If you receive a ciphertext file that has the signature certificate
192: glued to the message, you can still pry the signature certificate
193: away from the message during the decryption. You can do this with
194: the -b option during decrypt, like so:
195:
196: pgp -b letter
197:
198: This decrypts the letter.pgp file and if there is a signature in it,
199: PGP checks the signature and detaches it from the rest of the
200: message, storing it in the file letter.sig.
201:
202:
203: Decrypting the Message and Leaving the Signature on it
204: ------------------------------------------------------
205:
206: Usually, you want PGP to completely unravel a ciphertext file,
207: decrypting it and checking the nested signature if there is one,
208: peeling away the layers until you are left with only the original
209: plaintext file.
210:
211: But sometimes you want to decrypt an encrypted file, and leave the
212: inner signature still attached, so that you are left with a decrypted
213: signed message. This may be useful if you want to send a copy of a
214: signed document to a third party, perhaps re-enciphering it. For
215: example, suppose you get a message signed by Charlie, encrypted to
216: you. You want to decrypt it, and, leaving Charlie's signature on it,
217: you want to send it to Alice, perhaps re-enciphering it with Alice's
218: public key. No problem. PGP can handle that.
219:
220: To simply decrypt a message and leave the signature on it intact,
221: type:
222:
223: pgp -d letter
224:
225: This decrypts letter.pgp, and if there is an inner signature, it is
226: left intact with the decrypted plaintext in the output file.
227:
228: Now you can archive it, or maybe re-encrypt it and send it to someone
229: else.
230:
231:
232:
233: Sending ASCII Text Files Across Different Machine Environments
234: --------------------------------------------------------------
235:
236: You may use PGP to encrypt any kind of plaintext file, binary 8-bit
237: data or ASCII text. Probably the most common usage of PGP will be for
238: E-mail, when the plaintext is ASCII text.
239:
240: ASCII text is sometimes represented differently on different
241: machines. For example, on an MSDOS system, all lines of ASCII text
242: are terminated with a carriage return followed by a linefeed. On a
243: Unix system, all lines end with just a linefeed. On a Macintosh, all
244: lines end with just a carriage return. This is a sad fact of life.
245:
246: Normal unencrypted ASCII text messages are often automatically
247: translated to some common "canonical" form when they are transmitted
248: from one machine to another. Canonical text has a carriage return
249: and a linefeed at the end of each line of text. For example, the
250: popular KERMIT communication protocol can convert text to canonical
251: form when transmitting it to another system. This gets converted
252: back to local text line terminators by the receiving KERMIT. This
253: makes it easy to share text files across different systems.
254:
255: But encrypted text cannot be automatically converted by a
256: communication protocol, because the plaintext is hidden by
257: encipherment. To remedy this inconvenience, PGP lets you specify
258: that the plaintext should be treated as ASCII text (not binary data)
259: and should be converted to canonical text form before it gets
260: encrypted. At the receiving end, the decrypted plaintext is
261: automatically converted back to whatever text form is appropriate for
262: the local environment.
263:
264: To make PGP assume the plaintext is text that should be converted to
265: canonical text before encryption, just add the "t" option when
266: encrypting or signing a message, like so:
267:
268: pgp -et message.txt her_userid
269:
270: This mode is automatically turned off if PGP detects that the
271: plaintext file contains what it thinks is non-text binary data.
272:
273: For PGP users that use non-English 8-bit character sets, when PGP
274: converts text to canonical form, it may convert data from the local
275: character set into the LATIN1 (ISO 8859-1 Latin Alphabet 1) character
276: set, depending on the setting of the CHARSET parameter in the PGP
277: configuration file. LATIN1 is a superset of ASCII, with extra
278: characters added for many European languages.
279:
280:
281:
282: Leaving No Traces of Plaintext on the Disk
283: ------------------------------------------
284:
285: After PGP makes a ciphertext file for you, you can have PGP
286: automatically overwrite the plaintext file and delete it, leaving no
287: trace of plaintext on the disk so that no one can recover it later
288: using a disk block scanning utility. This is useful if the plaintext
289: file contains sensitive information that you don't want to keep
290: around.
291:
292: To wipe out the plaintext file after producing the ciphertext file,
293: just add the "w" (wipe) option when encrypting or signing a message,
294: like so:
295:
296: pgp -esw message.txt her_userid
297:
298: This example creates the ciphertext file "message.pgp", and the
299: plaintext file "message.txt" is destroyed beyond recovery.
300:
301: Obviously, you should be careful with this option. Also note that
302: this will not wipe out any fragments of plaintext that your word
303: processor might have created on the disk while you were editing the
304: message before running PGP. Most word processors create backup
305: files, scratch files, or both. Also, it overwrites the file only
306: once, which is enough to thwart conventional disk recovery efforts,
307: but not enough to withstand a determined and sophisticated effort to
308: recover the faint magnetic traces of the data using special disk
309: recovery hardware.
310:
311:
312:
313: Displaying Decrypted Plaintext on Your Screen
314: ---------------------------------------------
315:
316: To view the decrypted plaintext output on your screen (like the
317: Unix-style "more" command), without writing it to a file, use the -m
318: (more) option while decrypting:
319:
320: pgp -m ciphertextfile
321:
322: This displays the decrypted plaintext display on your screen one
323: screenful at a time.
324:
325:
326:
327: Making a Message For Her Eyes Only
328: ----------------------------------
329:
330: To specify that the recipient's decrypted plaintext will be shown
331: ONLY on her screen and cannot be saved to disk, add the -m option:
332:
333: pgp -sem message.txt her_userid
334:
335: Later, when the recipient decrypts the ciphertext with her secret key
336: and pass phrase, the plaintext will be displayed on her screen but
337: will not be saved to disk. The text will be displayed as it would if
338: she used the Unix "more" command, one screenful at a time. If she
339: wants to read the message again, she will have to decrypt the
340: ciphertext again.
341:
342: This feature is the safest way for you to prevent your sensitive
343: message from being inadvertently left on the recipient's disk. This
344: feature was added at the request of a user who wanted to send
345: intimate messages to his lover, but was afraid she might accidentally
346: leave the decrypted messages on her husband's computer.
347:
348:
349:
350: Preserving the Original Plaintext Filename
351: ------------------------------------------
352:
353: Normally, PGP names the decrypted plaintext output file with a name
354: similar to the input ciphertext filename, but dropping the
355: extension. Or, you can override that convention by specifying an
356: output plaintext filename on the command line with the -o option.
357: For most E-mail, this is a reasonable way to name the plaintext file,
358: because you get to decide its name when you decipher it, and your
359: typical E-mail messages often come from useless original plaintext
360: filenames like "to_phil.txt".
361:
362: But when PGP encrypts a plaintext file, it always saves the original
363: filename and attaches it to the plaintext before it compresses and
364: encrypts the plaintext. Normally, this hidden original filename is
365: discarded by PGP when it decrypts, but you can tell PGP you want to
366: preserve the original plaintext filename and use it as the name of
367: the decrypted plaintext output file. This is useful if PGP is used
368: to on files whose names are important to preserve.
369:
370: To recover the original plaintext filename while decrypting, add
371: the -p option, like so:
372:
373: pgp -p ciphertextfile
374:
375: I usually don't use this option, because if I did, about half of my
376: incoming E-mail would decrypt to the same plaintext filenames of
377: "to_phil.txt" or "prz.txt".
378:
379:
380:
381: Editing Your User ID or Pass Phrase
382: -----------------------------------
383:
384: Sometimes you may need to change your pass phrase, perhaps because
385: someone looked over your shoulder while you typed it in.
386:
387: Or you may need to change your user ID, because you got married and
388: changed your name, or maybe you changed your E-mail address. Or
389: maybe you want to add a second or third user ID to your key, because
390: you may be known by more than one name or E-mail address or job
391: title. PGP lets you attach more than one user ID to your key, any
392: one of which may be used to look up your key on the key ring.
393:
394: To edit your userid or pass phrase for your secret key:
395:
396: pgp -ke userid [keyring]
397:
398: PGP prompts you for a new user ID or a new pass phrase.
399:
400:
401:
402: Editing the Trust Parameters for a Public Key
403: ---------------------------------------------
404:
405: Sometimes you need to alter the trust parameters for a public key on
406: your public key ring. For a discussion on what these trust
407: parameters mean, see the section "How Does PGP Keep Track of Which
408: Keys are Valid?" in the Essential Topics volume of the PGP User's
409: Guide.
410:
411: To edit the trust parameters for a public key:
412:
413: pgp -ke userid [keyring]
414:
415:
416:
417: Checking If Everything is OK on Your Public Key Ring
418: ----------------------------------------------------
419:
420: Normally, PGP automatically checks any new keys or signatures on your
421: public key ring and updates all the trust parameters and validity
422: scores. In theory, it keeps all the key validity status information
423: up to date as material is added to or deleted from your public key
424: ring. But perhaps you may want to explicitly force PGP to perform a
425: comprehensive analysis of your public key ring, checking all the
426: certifying signatures, checking the trust parameters, updating all
427: the validity scores, and checking your own ultimately-trusted key
428: against a backup copy on a write-protected floppy disk. It may be a
429: good idea to do this hygienic maintenance periodically to make sure
430: nothing is wrong with your public key ring. To force PGP to perform
431: a full analysis of your public key ring, use the -kc (key ring check)
432: command:
433:
434: pgp -kc
435:
436: You can also make PGP check all the signatures for just a single
437: selected public key by:
438:
439: pgp -kc userid [keyring]
440:
441: For further information on how the backup copy of your own key is
442: checked, see the description of the BAKRING parameter in the
443: configuration file section of this manual.
444:
445:
446:
447: Verifying a Public Key Over the Phone
448: -------------------------------------
449:
450: If you get a public key from someone that is not certified by anyone
451: you trust, how can you tell if it's really their key? The best way
452: to verify an uncertified key is to verify it over some independent
453: channel other than the one you received the key through. One
454: convenient way to tell, if you know this person and would recognize
455: them on the phone, is to call them and verify their key over the
456: telephone. Rather than reading their whole tiresome (ASCII-armored)
457: key to them over the phone, you can just read their key's
458: "fingerprint" to them. To see this fingerprint, use the -kvc
459: command:
460:
461: pgp -kvc userid [keyring]
462:
463: This will display the key with the 16-byte digest of the public key
464: components. Read this 16-byte fingerprint to the key's owner on the
465: phone, while she checks it against her own, using the same -kvc
466: command at her end.
467:
468: You can both verify each other's keys this way, and then you can sign
469: each other's keys with confidence. This is a safe and convenient way
470: to get the key trust network started for your circle of friends.
471:
472:
473:
474: Using PGP as a Unix-style Filter
475: --------------------------------
476:
477: Unix fans are accustomed to using Unix "pipes" to make two
478: applications work together. The output of one application can be
479: directly fed through a pipe to be read as input to another
480: application. For this to work, the applications must be capable of
481: reading the raw material from "standard input" and writing the
482: finished output to "standard output". PGP can operate in this mode.
483: If you don't understand what this means, then you probably don't need
484: this feature.
485:
486: To use a Unix-style filter mode, reading from standard input and
487: writing to standard output, add the -f option, like so:
488:
489: pgp -feast her_userid <inputfile >outputfile
490:
491: This feature makes it easier to make PGP work with electronic mail
492: applications.
493:
494: When using PGP in filter mode to decrypt a ciphertext file, you may
495: find it useful to use the PGPPASS environmental variable to hold the
496: pass phrase, so that you won't be prompted for it. The PGPPASS
497: feature is explained below.
498:
499:
500:
501: Suppressing Unneccessary Questions: BATCHMODE
502: ----------------------------------------------
503:
504: With the BATCHMODE flag enabled on the command line, PGP will not ask
505: any unneccessary questions or prompt for alternate filenames. Here
506: is an example of how to set this flag:
507:
508: pgp +batchmode cipherfile
509:
510: This is useful for running PGP non-interactively from Unix shell
511: scripts or MSDOS batch files. Some key management commands still
512: need user interaction even when BATCHMODE is on, so shell scripts may
513: need to avoid them.
514:
515: BATCHMODE may also be enabled to check the validity of a signature on
516: a file. If there was no signature on the file, the exit code is 1.
517: If it had a signature that was good, the exit code is 0.
518:
519:
520: Force "Yes" Answer to Confirmation Questions: FORCE
521: ----------------------------------------------------
522:
523: This command-line flag makes PGP assume "yes" for the user response
524: to the confirmation request to overwrite an existing file, or when
525: removing a key from the keyring via the -kr command. Here is an
526: example of how to set this flag:
527:
528: pgp +force cipherfile
529: or:
530: pgp -kr +force Smith
531:
532: This feature is useful for running PGP non-interactively from a Unix
533: shell script or MSDOS batch file.
534:
535:
536:
537: PGP Returns Exit Status to the Shell
538: ------------------------------------
539:
540: To facilitate running PGP in "batch" mode, such as from an MSDOS
541: ".bat" file or from a Unix shell script, PGP returns an error exit
542: status to the shell. An exit status code of zero means normal exit,
543: while a nonzero exit status indicates some kind of error occurred.
544: Different error exit conditions return different exit status codes to
545: the shell.
546:
547:
548:
549: Environmental Variable for Pass Phrase
550: --------------------------------------
551:
552: Normally, PGP prompts the user to type a pass phrase whenever PGP
553: needs a pass phrase to unlock a secret key. But it is possible to
554: store the pass phrase in an environmental variable from your
555: operating system's command shell. The environmental variable PGPPASS
556: can be used to hold the pass phrase that PGP will attempt to use
557: first. If the pass phrase stored in PGPPASS is incorrect, PGP
558: recovers by prompting the user for the correct pass phrase.
559:
560: For example, on MSDOS, the shell command:
561:
562: SET PGPPASS=zaphod beeblebrox for president
563:
564: would eliminate the prompt for the pass phrase if the pass phrase
565: were indeed "zaphod beeblebrox for president".
566:
567: This dangerous feature makes your life more convenient if you have to
568: regularly deal with a large number of incoming messages addressed to
569: your secret key, by eliminating the need for you to repeatedly type
570: in your pass phrase every time you run PGP.
571:
572: I added this feature because of popular demand. However, this is a
573: somewhat dangerous feature, because it keeps your precious pass
574: phrase stored somewhere other than just in your brain. Even worse,
575: if you are particularly reckless, it may even be stored on a disk on
576: the same computer as your secret key. It would be particularly
577: dangerous and stupid if you were to install this command in a batch
578: or script file, such as the MSDOS AUTOEXEC.BAT file. Someone could
579: come along on your lunch hour and steal both your secret key ring and
580: the file containing your pass phrase.
581:
582: I can't emphasize the importance of this risk enough. If you are
583: contemplating using this feature, be sure to read the sections
584: "Exposure on Multi-user Systems" and "How to Protect Secret Keys from
585: Disclosure" in this volume and in the Essential Topics volume of the
586: PGP User's Guide.
587:
588: If you must use this feature, the safest way to do it would be to
589: just manually type in the shell command to set PGPPASS every time you
590: boot your machine to start using PGP, and then erase it or turn off
591: your machine when you are done. And you should definitely never do
592: it in an environment where someone else may have access to your
593: machine. Someone could come along and simply ask your computer to
594: display the contents of PGPPASS.
595:
596:
597:
598: Setting Configuration Parameters: CONFIG.TXT
599: ============================================
600:
601: PGP has a number of user-settable parameters that can be defined in a
602: special configuration text file called "config.txt", in the directory
603: pointed to by the shell environmental variable PGPPATH. Having a
604: configuration file enables the user to define various flags and
605: parameters for PGP without the burden of having to always define
606: these parameters in the PGP command line.
607:
608: Configuration parameters may be assigned integer values, character
609: string values, or on/off values, depending on what kind of
610: configuration parameter it is. A sample configuration file is
611: provided with PGP, so you can see some examples.
612:
613: In the configuration file, blank lines are ignored, as is anything
614: following the '#' comment character. Keywords are not
615: case-sensitive.
616:
617: Here is a short sample fragment of a typical configuration file:
618:
619: # TMP is the directory for PGP scratch files, such as a RAM disk.
620: TMP = "e:\" # Can be overridden by environment variable TMP.
621: Armor = on # Use -a flag for ASCII armor whenever applicable.
622: # CERT_DEPTH is how deeply introducers may introduce introducers.
623: cert_depth = 3
624:
625: If some configuration parameters are not defined in the configuration
626: file, or if there is no configuration file, or if PGP can't find the
627: configuration file, the values for the configuration parameters
628: default to some reasonable value.
629:
630: Note that it is also possible to set these same configuration
631: parameters directly from the PGP command line, by preceding the
632: parameter setting with a "+" character. For example, the following
633: two PGP commands produce the same effect:
634:
635: pgp -e +armor=on message.txt smith
636: or: pgp -ea message.txt smith
637:
638:
639: The following is a summary of the various parameters than may be
640: defined in the configuration file.
641:
642:
643: TMP - Directory Pathname for Temporary Files
644: --------------------------------------------
645:
646: Default setting: TMP = ""
647:
648: The configuration parameter TMP specifies what directory to use for
649: PGP's temporary scratch files. The best place to put them is on a
650: RAM disk, if you have one. That speeds things up quite a bit, and
651: increases security somewhat. If TMP is undefined, the temporary
652: files go in the current directory. If the shell environmental
653: variable TMP is defined, PGP instead uses that to specify where the
654: temporary files should go.
655:
656:
657: LANGUAGE - Foreign Language Selector
658: ------------------------------------
659:
660: Default setting: LANGUAGE = "en"
661:
662: PGP displays various prompts, warning messages, and advisories to the
663: user on the screen. For example, messages such as "File not found.",
664: or "Please enter your pass phrase:". These messages are normally in
665: English. But it is possible to get PGP to display its messages to
666: the user in other languages, without having to modify the PGP
667: executable program.
668:
669: A number of people in various countries have translated all of PGP's
670: display messages, warnings, and prompts into their native languages.
671: These hundreds of translated message strings have been placed in a
672: special text file called "language.txt", distributed with the PGP
673: release. The messages are stored in this file in English, Spanish,
674: Dutch, German, French, Italian, Russian, Latvian, and Lithuanian.
675: Other languages may be added later.
676:
677: The configuration parameter LANGUAGE specifies what language to
678: display these messages in. LANGUAGE may be set to "en" for English,
679: "es" for Spanish, "de" for German, "nl" for Dutch, "fr" for French,
680: "it" for Italian, "ru" for Russian, "lt3" for Lithuanian, "lv" for
681: Latvian, "esp" for Esperanto. For example, if this line appeared in
682: the configuration file:
683:
684: LANGUAGE = "fr"
685:
686: PGP would select French as the language for its display messages.
687: The default setting is English.
688:
689: When PGP needs to display a message to the user, it looks in the
690: "language.txt" file for the equivalent message string in the selected
691: foreign language and displays that translated message to the user.
692: If PGP can't find the language string file, or if the selected
693: language is not in the file, or if that one phrase is not translated
694: into the selected language in the file, or if that phrase is missing
695: entirely from the file, PGP displays the message in English.
696:
697: To conserve disk space, most foreign translations are not included
698: in the standard PGP release package, but are available separately.
699:
700:
701: MYNAME - Default User ID for Making Signatures
702: ----------------------------------------------
703:
704: Default setting: MYNAME = ""
705:
706: The configuration parameter MYNAME specifies the default user ID to
707: use to select the secret key for making signatures. If MYNAME is not
708: defined, the most recent secret key you installed on your secret key
709: ring will be used. The user may also override this setting by
710: specifying a user ID on the PGP command line with the -u option.
711:
712:
713: TEXTMODE - Assuming Plaintext is a Text File
714: --------------------------------------------
715:
716: Default setting: TEXTMODE = off
717:
718: The configuration parameter TEXTMODE is equivalent to the -t command
719: line option. If enabled, it causes PGP to assume the plaintext is a
720: text file, not a binary file, and converts it to "canonical text"
721: before encrypting it. Canonical text has a carriage return and a
722: linefeed at the end of each line of text.
723:
724: This mode will be automatically turned off if PGP detects that the
725: plaintext file contains what it thinks is non-text binary data.
726:
727: For VAX/VMS systems, the current version of PGP defaults TEXTMODE=ON.
728:
729: For further details, see the section "Sending ASCII Text Files Across
730: Different Machine Environments".
731:
732:
733: CHARSET - Specifies Local Character Set for Text Files
734: ------------------------------------------------------
735:
736: Default setting: CHARSET = NOCONV
737:
738: Because PGP must process messages in many non-English languages with
739: non-ASCII character sets, you may have a need to tell PGP what local
740: character set your machine uses. This determines what character
741: conversions are performed when converting plaintext files to and from
742: canonical text format. This is only a concern if you are in a
743: non-English non-ASCII environment.
744:
745: The configuration parameter CHARSET selects the local character set.
746: The choices are NOCONV (no conversion), LATIN1 (ISO 8859-1 Latin
747: Alphabet 1), KOI8 (used by most Russian Unix systems), ALT_CODES
748: (used by Russian MSDOS systems), ASCII, and CP850 (used by most
749: western European languages on standard MSDOS PCs).
750:
751: LATIN1 is the internal representation used by PGP for canonical text,
752: so if you select LATIN1, no conversion is done. Note also that PGP
753: treats KOI8 as LATIN1, even though it is a completely different
754: character set (Russian), because trying to convert KOI8 to either
755: LATIN1 or CP850 would be futile anyway. This means that setting
756: CHARSET to NOCONV, LATIN1, or KOI8 are all equivalent to PGP.
757:
758: If you use MSDOS and expect to send or receive traffic in western
759: European languages, set CHARSET = "CP850". This will make PGP
760: convert incoming canonical text messages from LATIN1 to CP850 after
761: decryption. If you use the -t (textmode) option to convert to
762: canonical text, PGP will convert your CP850 text to LATIN1 before
763: encrypting it.
764:
765: For further details, see the section "Sending ASCII Text Files Across
766: Different Machine Environments".
767:
768:
769: ARMOR - Enable ASCII Armor Output
770: ---------------------------------
771:
772: Default setting: ARMOR = off
773:
774: The configuration parameter ARMOR is equivalent to the -a command
775: line option. If enabled, it causes PGP to emit ciphertext or keys in
776: ASCII Radix-64 format suitable for transporting through E-mail
777: channels. Output files are named with the ".asc" extension.
778:
779: If you tend to use PGP mostly for E-mail, it may be a good idea to
780: enable this parameter.
781:
782: For further details, see the section "Sending Ciphertext Through
783: E-mail Channels: Radix-64 Format" in the Essential Topics volume.
784:
785:
786: ARMORLINES - Size of ASCII Armor Multipart Files
787: ------------------------------------------------
788:
789: Default setting: ARMORLINES = 720
790:
791: When PGP creates a very large ".asc" radix-64 file for sending
792: ciphertext or keys through the E-mail, it breaks the file up into
793: separate chunks small enough to send through Internet mail
794: utilities. Normally, Internet mailers prohibit files larger than
795: about 50000 bytes, which means that if we restrict the number of
796: lines to about 720, we'll be well within the limit. The file chunks
797: are named with suffixes ".as1", ".as2", ".as3", ...
798:
799: The configuration parameter ARMORLINES specifies the maximum number
800: of lines to make each chunk in a multipart ".asc" file sequence. If
801: you set it to zero, PGP will not break up the file into chunks.
802:
803: Fidonet email files usually have an upper limit of about 32K bytes,
804: so 450 lines would be appropriate for Fidonet environments.
805:
806: For further details, see the section "Sending Ciphertext Through
807: E-mail Channels: Radix-64 Format" in the Essential Topics volume.
808:
809:
810: KEEPBINARY - Keep Binary Ciphertext Files After Decrypting
811: ----------------------------------------------------------
812:
813: Default setting: KEEPBINARY = off
814:
815: When PGP reads a ".asc" file, it recognizes that the file is in
816: radix-64 format and will convert it back to binary before processing
817: as it normally does, producing as a by-product a ".pgp" ciphertext
818: file in binary form. After further processing to decrypt the ".pgp"
819: file, the final output file will be in normal plaintext form.
820:
821: You may want to delete the binary ".pgp" intermediate file, or you
822: may want PGP to delete it for you automatically. You can still rerun
823: PGP on the original ".asc" file.
824:
825: The configuration parameter KEEPBINARY enables or disables keeping
826: the intermediate ".pgp" file during decryption.
827:
828: For further details, see the section "Sending Ciphertext Through
829: E-mail Channels: Radix-64 Format" in the Essential Topics volume.
830:
831:
832: COMPRESS - Enable Compression
833: -----------------------------
834:
835: Default setting: COMPRESS = on
836:
837: The configuration parameter COMPRESS enables or disables data
838: compression before encryption. It is used mainly for debugging PGP.
839: Normally, PGP attempts to compress the plaintext before it encrypts
840: it. Generally, you should leave this alone and let PGP attempt to
841: compress the plaintext.
842:
843:
844: COMPLETES_NEEDED - Number of Completely Trusted Introducers Needed
845: ------------------------------------------------------------------
846:
847: Default setting: COMPLETES_NEEDED = 1
848:
849: The configuration parameter COMPLETES_NEEDED specifies the minimum
850: number of completely trusted introducers required to fully certify a
851: public key on your public key ring. This gives you a way of tuning
852: PGP's skepticism.
853:
854: For further details, see the section "How Does PGP Keep Track of
855: Which Keys are Valid?" in the Essential Topics volume.
856:
857:
858: MARGINALS_NEEDED - Number of Marginally Trusted Introducers Needed
859: ------------------------------------------------------------------
860:
861: Default setting: MARGINALS_NEEDED = 2
862:
863: The configuration parameter MARGINALS_NEEDED specifies the minimum
864: number of marginally trusted introducers required to fully certify a
865: public key on your public key ring. This gives you a way of tuning
866: PGP's skepticism.
867:
868: For further details, see the section "How Does PGP Keep Track of
869: Which Keys are Valid?" in the Essential Topics volume.
870:
871:
872: CERT_DEPTH - How Deep May Introducers Be Nested
873: -----------------------------------------------
874:
875: Default setting: CERT_DEPTH = 4
876:
877: The configuration parameter CERT_DEPTH specifies how many levels deep
878: you may nest introducers to certify other introducers to certify
879: public keys on your public key ring. For example, If CERT_DEPTH is
880: set to 1, there may only be one layer of introducers below your own
881: ultimately-trusted key. If that were the case, you would be required
882: to directly certify the public keys of all trusted introducers on
883: your key ring. If you set CERT_DEPTH to 0, you could have no
884: introducers at all, and you would have to directly certify each and
885: every key on your public key ring in order to use it. The minimum
886: CERT_DEPTH is 0, the maximum is 8.
887:
888: For further details, see the section "How Does PGP Keep Track of
889: Which Keys are Valid?" in the Essential Topics volume.
890:
891:
892: BAKRING - Filename for Backup Secret Keyring
893: --------------------------------------------
894:
895: Default setting: BAKRING = ""
896:
897: All of the key certification that PGP does on your public key ring
898: ultimately depends on your own ultimately-trusted public key (or
899: keys). To detect any tampering of your public key ring, PGP must
900: check that your own key has not been tampered with. To do this, PGP
901: must compare your public key against a backup copy of your secret key
902: on some tamper-resistant media, such as a write-protected floppy
903: disk. A secret key contains all the information that your public key
904: has, plus some secret components. This means PGP can check your
905: public key against a backup copy of your secret key.
906:
907: The configuration parameter BAKRING specifies what pathname to use
908: for PGP's trusted backup copy of your secret key ring. On MSDOS, you
909: could set it to "a:\secring.pgp" to point it at a write-protected
910: backup copy of your secret key ring on your floppy drive. This check
911: is performed only when you execute the PGP -kc option to check your
912: whole public key ring.
913:
914: If BAKRING is not defined, PGP will not check your own key against
915: any backup copy.
916:
917: For further details, see the sections "How to Protect Public Keys
918: from Tampering" and "How Does PGP Keep Track of Which Keys are
919: Valid?" in the Essential Topics volume.
920:
921:
922: PAGER - Selects Shell Command to Display Plaintext Output
923: ---------------------------------------------------------
924:
925: Default setting: PAGER = ""
926:
927: PGP lets you view the decrypted plaintext output on your screen (like
928: the Unix-style "more" command), without writing it to a file, if you
929: use the -m (more) option while decrypting. This displays the
930: decrypted plaintext display on your screen one screenful at a time.
931:
932: If you prefer to use a fancier page display utility, rather than
933: PGP's built-in one, you can specify the name of a shell command that
934: PGP will invoke to display your plaintext output file. The
935: configuration parameter PAGER specifies the shell command to invoke
936: to display the file. For example, on MSDOS systems, you might want
937: to use the popular shareware program "list.com" to display your
938: plaintext message. Assuming you have a copy of "list.com", you may
939: set PAGER accordingly:
940:
941: PAGER = "list"
942:
943: However, if the sender specified that this file is for your eyes
944: only, and may not be written to disk, PGP always uses its own
945: built-in display function.
946:
947: For further details, see the section "Displaying Decrypted Plaintext
948: on Your Screen".
949:
950:
951: SHOWPASS - Echo Pass Phrase to User
952: -----------------------------------
953:
954: Default setting: SHOWPASS = off
955:
956: Normally, PGP does not let you see your pass phrase as you type it
957: in. This makes it harder for someone to look over your shoulder
958: while you type and learn your pass phrase. But some typing-impaired
959: people have problems typing their pass phrase without seeing what
960: they are typing, and they may be typing in the privacy of their own
961: homes. So they asked if PGP can be configured to let them see what
962: they type when they type in their pass phrase.
963:
964: The configuration parameter SHOWPASS enables PGP to echo your typing
965: during pass phrase entry.
966:
967:
968: TZFIX - Timezone Adjustment
969: ---------------------------
970:
971: Default setting: TZFIX = 0
972:
973: PGP provides timestamps for keys and signature certificates in
974: Greenwich Mean Time (GMT), or Coordinated Universal Time (UTC), which
975: means the same thing for our purposes. When PGP asks the system for
976: the time of day, the system is supposed to provide it in GMT.
977:
978: But sometimes, because of improperly configured MSDOS systems, the
979: system time is returned in US Pacific Standard Time time plus 8
980: hours. Sounds weird, doesn't it? Perhaps because of some sort of US
981: west-coast jingoism, MSDOS presumes local time is US Pacific time,
982: and pre-corrects Pacific time to GMT. This adversely affects the
983: behavior of the internal MSDOS GMT time function that PGP calls.
984: However, if your MSDOS environmental variable TZ is already properly
985: defined for your timezone, this corrects the misconception MSDOS has
986: that the whole world lives on the US west coast.
987:
988: The configuration parameter TZFIX specifies the number of hours to
989: add to the system time function to get GMT, for GMT timestamps on
990: keys and signatures. If the MSDOS environmental variable TZ is
991: defined properly, you can leave TZFIX=0. Unix systems usually
992: shouldn't need to worry about setting TZFIX at all. But if you are
993: using some other obscure operating system that doesn't know about
994: GMT, you may have to use TZFIX to adjust the system time to GMT.
995:
996: On MSDOS systems that do not have TZ defined in the environment, you
997: should make TZFIX=0 for California, -1 for Colorado, -2 for Chicago,
998: -3 for New York, -8 for London, -9 for Amsterdam. In the summer,
999: TZFIX should be manually decremented from these values. What a mess.
1000:
1001: It would be much cleaner to set your MSDOS environmental variable TZ
1002: in your AUTOEXEC.BAT file, and not use the TZFIX correction. Then
1003: MSDOS gives you good GMT timestamps, and will handle daylight savings
1004: time adjustments for you. Here are some sample lines to insert into
1005: AUTOEXEC.BAT, depending on your time zone:
1006:
1007: For Los Angeles: SET TZ=PST8PDT
1008: For Denver: SET TZ=MST7MDT
1009: For Arizona: SET TZ=MST7
1010: (Arizona never uses daylight savings time)
1011: For Chicago: SET TZ=CST6CDT
1012: For New York: SET TZ=EST5EDT
1013: For London: SET TZ=GMT0BST
1014: For Amsterdam: SET TZ=MET-1DST
1015: For Moscow: SET TZ=MSK-3MSD
1016: For Aukland: SET TZ=NZT-13
1017:
1018:
1019: CLEARSIG - Enable Signed Messages to be Encapsulated as Clear Text
1020: ------------------------------------------------------------------
1021:
1022: Default setting: CLEARSIG = off
1023:
1024: Normally, unencrypted PGP signed messages have a signature
1025: certificate prepended in binary form. To send this through a 7-bit
1026: E-mail channel, radix-64 ASCII armor is applied (see the ARMOR
1027: parameter), rendering the message unreadable to casual human eyes,
1028: even though the message is not actually encrypted. The recipient
1029: must use PGP to strip the armor off before reading the message.
1030:
1031: If the original plaintext message is in text (not binary) form, there
1032: is a way to send it through an E-mail channel in such a way that the
1033: ASCII armor is applied only to the binary signature certificate, but
1034: not to the plaintext message. This makes it possible to read the
1035: signed message with human eyes, without the aid of PGP. Of course,
1036: you still need PGP to actually check the signature.
1037:
1038: To enable this feature, set CLEARSIG=ON, and set ARMOR=ON (or use
1039: the -a option), and set TEXTMODE=ON (or use the -t option). For
1040: example, you can set CLEARSIG directly from the command line:
1041:
1042: pgp -sta +clearsig=on message.txt
1043:
1044: This message representation is analogous to the MIC-CLEAR message type
1045: used in Internet Privacy Enhanced Mail (PEM). It is important to
1046: note that since this method only applies ASCII armor to the binary
1047: signature certificate, and not to the message text itself, there is
1048: some risk that the unarmored message may suffer some accidental
1049: molestation while en route. This can happen if it passes through
1050: some E-mail gateway that performs character set conversions, or in
1051: some cases extra spaces may be added to or stripped from the ends of
1052: lines. If this occurs, the signature will fail to verify, which may
1053: give a false indication of intentional tampering. But since PEM
1054: lives under a similar vulnerability, it seems worth having this
1055: feature despite the risks.
1056:
1057: Beginning with PGP version 2.2, trailing blanks are ignored on each
1058: line in calculating the signature for text in CLEARSIG mode.
1059:
1060:
1061: VERBOSE - Quiet, Normal, or Verbose Messages
1062: --------------------------------------------
1063:
1064: Default setting: VERBOSE = 1
1065:
1066: VERBOSE may be set to 0, 1, or 2, depending on how much detail you
1067: want to see from PGP diagnostic messages. The settings are:
1068:
1069: 0 - Display messages only if there is a problem. Unix fans wanted
1070: this "quiet mode" setting.
1071:
1072: 1 - Normal default setting. Displays a reasonable amount of detail
1073: in diagnostic or advisory messages.
1074:
1075: 2 - Displays maximum information, usually to help diagnose problems
1076: in PGP. Not recommended for normal use. Besides, PGP doesn't have
1077: any problems, right?
1078:
1079:
1080: INTERACTIVE - Ask for Confirmation for Key Adds
1081: -----------------------------------------------
1082:
1083: Default Setting: INTERACTIVE = off
1084:
1085: Enabling this mode will mean that if you add a key file containing
1086: multiple keys to your key ring, PGP will ask for confirmation for
1087: each key before adding it to your key ring.
1088:
1089:
1090:
1091: Protecting Against Bogus Timestamps
1092: ===================================
1093:
1094: A somewhat obscure vulnerability of PGP involves dishonest users
1095: creating bogus timestamps on their own public key certificates and
1096: signatures. You can skip over this section if you are a casual user
1097: and aren't deeply into obscure public key protocols.
1098:
1099: There's nothing to stop a dishonest user from altering the date and
1100: time setting of his own system's clock, and generating his own public
1101: key certificates and signatures that appear to have been created at a
1102: different time. He can make it appear that he signed something
1103: earlier or later than he actually did, or that his public/secret key
1104: pair was created earlier or later. This may have some legal or
1105: financial benefit to him, for example by creating some kind of
1106: loophole that might allow him to repudiate a signature.
1107:
1108: A remedy for this could involve some trustworthy Certifying Authority
1109: or notary that would create notarized signatures with a trustworthy
1110: timestamp. This might not necessarily require a centralized
1111: authority. Perhaps any trusted introducer or disinterested party
1112: could serve this function, the same way real notary publics do now.
1113: A public key certificate could be signed by the notary, and the
1114: trusted timestamp in the notary's signature would have some legal
1115: significance. The notary could enter the signed certificate into a
1116: special certificate log controlled by the notary. Anyone can read
1117: this log.
1118:
1119: The notary could also sign other people's signatures, creating a
1120: signature certificate of a signature certificate. This would serve
1121: as a witness to the signature the same way real notaries do now with
1122: paper. Again, the notary could enter the detached signature
1123: certificate (without the actual whole document that was signed) into
1124: a log controlled by the notary. The notary's signature would have a
1125: trusted timestamp, which might have greater credibility than the
1126: timestamp in the original signature. A signature becomes "legal" if
1127: it is signed and logged by the notary.
1128:
1129: This problem of certifying signatures with notaries and trusted
1130: timestamps warrants further discussion. This can of worms will not
1131: be fully covered here now. There is a good treatment of this topic
1132: in Denning's 1983 article in IEEE Computer (see references). There
1133: is much more detail to be worked out in these various certifying
1134: schemes. This will develop further as PGP usage increases and other
1135: public key products develop their own certifying schemes.
1136:
1137:
1138: A Peek Under the Hood
1139: =====================
1140:
1141: Let's take a look at a few internal features of PGP.
1142:
1143:
1144: Random Numbers
1145: --------------
1146:
1147: PGP uses a cryptographically strong pseudorandom number generator for
1148: creating temporary conventional session keys. The seed file for this
1149: is called "randseed.bin". It too can be kept in whatever directory
1150: is indicated by the PGPPATH environmental variable. If this random
1151: seed file does not exist, it is automatically created and seeded with
1152: truly random numbers derived from timing your keystroke latencies.
1153:
1154: This generator reseeds the disk file each time it is used by mixing
1155: in new key material partially derived with the time of day and other
1156: truly random sources. It uses the conventional encryption algorithm
1157: as an engine for the random number generator. The seed file contains
1158: both random seed material and random key material to key the
1159: conventional encryption engine for the random generator.
1160:
1161: This random seed file should be at least slightly protected from
1162: disclosure, to reduce the risk of an attacker deriving your next or
1163: previous session keys. The attacker would have a very hard time
1164: getting anything useful from capturing this random seed file, because
1165: the file is cryptographically laundered before and after each use.
1166: Nonetheless, it seems prudent to at least try to keep it from falling
1167: into the wrong hands.
1168:
1169: If you feel uneasy about trusting any algorithmically derived random
1170: number source however strong, keep in mind that you already trust the
1171: strength of the same conventional cipher to protect your messages.
1172: If it's strong enough for that, then it should be strong enough to
1173: use as a source of random numbers for temporary session keys. Note
1174: that PGP still uses truly random numbers from physical sources
1175: (mainly keyboard timings) to generate long-term public/secret key
1176: pairs.
1177:
1178:
1179:
1180: PGP's Conventional Encryption Algorithm
1181: ---------------------------------------
1182:
1183: As described earlier, PGP "bootstraps" into a conventional single-key
1184: encryption algorithm by using a public key algorithm to encipher the
1185: conventional session key and then switching to fast conventional
1186: cryptography. So let's talk about this conventional encryption
1187: algorithm. It isn't the DES.
1188:
1189: The Federal Data Encryption Standard (DES) is a good algorithm for
1190: most commercial applications. However, the Government does not trust
1191: the DES to protect its own classified data. Perhaps this is because
1192: the DES key length is 56 bits, short enough for a brute force attack
1193: with a special purpose machine built from massive numbers of DES
1194: chips. Also, Biham and Shamir have had some success recently on
1195: attacking the full 16-round DES.
1196:
1197: PGP does not use the DES as its conventional single-key algorithm to
1198: encrypt messages. Instead, PGP uses a different conventional
1199: single-key block encryption algorithm, called IDEA(tm). A future
1200: version of PGP may support the DES as an option, if enough users
1201: ask for it. But I suspect IDEA is better than DES.
1202:
1203: For the cryptographically curious, the IDEA cipher has a 64-bit block
1204: size for the plaintext and the ciphertext. It uses a key size of 128
1205: bits. It is based on the design concept of "mixing operations from
1206: different algebraic groups". It runs much faster in software than
1207: the DES. Like the DES, it can be used in cipher feedback (CFB) and
1208: cipher block chaining (CBC) modes. PGP uses it in 64-bit CFB mode.
1209:
1210: The IPES/IDEA block cipher was developed at ETH in Zurich by James L.
1211: Massey and Xuejia Lai, and published in 1990. This is not a
1212: "home-grown" algorithm. Its designers have a distinguished
1213: reputation in the cryptologic community. Early published papers on
1214: the algorithm called it IPES (Improved Proposed Encryption Standard),
1215: but they later changed the name to IDEA (International Data
1216: Encryption Algorithm). So far, IDEA has resisted attack much better
1217: than other ciphers such as FEAL, REDOC-II, LOKI, Snefru and Khafre.
1218: And preliminary evidence suggests that IDEA may be more resistant
1219: than the DES to Biham & Shamir's highly successful differential
1220: cryptanalysis attack. Biham and Shamir have been examining the IDEA
1221: cipher for weaknesses, without success. Academic cryptanalyst groups
1222: in Belgium, England, and Germany are also attempting to attack it, as
1223: well as the military services from several European countries. As
1224: this new cipher continues to attract attack efforts from the most
1225: formidable quarters of the cryptanalytic world, confidence in IDEA is
1226: growing with the passage of time.
1227:
1228: A famous hockey player once said, "I try to skate to where I think
1229: the puck will be." A lot of people are starting to feel that the
1230: days are numbered for the DES. I'm skating toward IDEA.
1231:
1232: It is not ergonomically practical to use pure RSA with large keys to
1233: encrypt and decrypt long messages. Absolutely no one does it that way
1234: in the real world. But perhaps you are concerned that the whole
1235: package is weakened if we use a hybrid public-key and conventional
1236: scheme just to speed things up. After all, a chain is only as strong
1237: as its weakest link. Many people less experienced in cryptography
1238: mistakenly believe that RSA is intrinsically stronger than any
1239: conventional cipher. It's not. RSA can be made weak by using weak
1240: keys, and conventional ciphers can be made strong by choosing good
1241: algorithms. It's usually difficult to tell exactly how strong a good
1242: conventional cipher is, without actually cracking it. A really good
1243: conventional cipher might possibly be harder to crack than even a
1244: "military grade" RSA key. The attraction of public key cryptography
1245: is not because it is intrinsically stronger than a conventional
1246: cipher-- its appeal is because it helps you manage keys more
1247: conveniently.
1248:
1249:
1250:
1251: Data Compression
1252: ----------------
1253:
1254: PGP normally compresses the plaintext before encrypting it. It's too
1255: late to compress it after it has been encrypted; encrypted data is
1256: incompressible. Data compression saves modem transmission time and
1257: disk space and more importantly strengthens cryptographic security.
1258: Most cryptanalysis techniques exploit redundancies found in the
1259: plaintext to crack the cipher. Data compression reduces this
1260: redundancy in the plaintext, thereby greatly enhancing resistance to
1261: cryptanalysis. It takes extra time to compress the plaintext, but
1262: from a security point of view it seems worth it, at least in my
1263: cautious opinion.
1264:
1265: Files that are too short to compress or just don't compress well are
1266: not compressed by PGP.
1267:
1268: If you prefer, you can use PKZIP to compress the plaintext before
1269: encrypting it. PKZIP is a widely-available and effective MSDOS
1270: shareware compression utility from PKWare, Inc. Or you can use ZIP,
1271: a PKZIP-compatible freeware compression utility on Unix and other
1272: systems, available from Jean-Loup Gailly. There is some advantage in
1273: using PKZIP or ZIP in certain cases, because unlike PGP's built-in
1274: compression algorithm, PKZIP and ZIP have the nice feature of
1275: compressing multiple files into a single compressed file, which is
1276: reconstituted again into separate files when decompressed. PGP will
1277: not try to compress a plaintext file that has already been
1278: compressed. After decrypting, the recipient can decompress the
1279: plaintext with PKUNZIP. If the decrypted plaintext is a PKZIP
1280: compressed file, PGP automatically recognizes this and advises the
1281: recipient that the decrypted plaintext appears to be a PKZIP file.
1282:
1283: For the technically curious readers, the current version of PGP uses
1284: the freeware ZIP compression routines written by Jean-loup Gailly,
1285: Mark Adler, and Richard B. Wales. This ZIP software uses
1286: functionally-equivalent compression algorithms as those used by
1287: PKWare's new PKZIP 2.0. This ZIP compression software was selected
1288: for PGP mainly because of its free portable C source code
1289: availability, and because it has a really good compression ratio, and
1290: because it's fast.
1291:
1292: Peter Gutmann has also written a nice compression utility called
1293: HPACK, available for free from many Internet FTP sites. It encrypts
1294: the compressed archives, using PGP data formats and key rings. He
1295: wanted me to mention that here.
1296:
1297:
1298:
1299: Message Digests and Digital Signatures
1300: --------------------------------------
1301:
1302: To create a digital signature, PGP encrypts with your secret key.
1303: But PGP doesn't actually encrypt your entire message with your secret
1304: key-- that would take too long. Instead, PGP encrypts a "message
1305: digest".
1306:
1307: The message digest is a compact (128 bit) "distillate" of your
1308: message, similar in concept to a checksum. You can also think of it
1309: as a "fingerprint" of the message. The message digest "represents"
1310: your message, such that if the message were altered in any way, a
1311: different message digest would be computed from it. This makes it
1312: possible to detect any changes made to the message by a forger. A
1313: message digest is computed using a cryptographically strong one-way
1314: hash function of the message. It would be computationally infeasible
1315: for an attacker to devise a substitute message that would produce an
1316: identical message digest. In that respect, a message digest is much
1317: better than a checksum, because it is easy to devise a different
1318: message that would produce the same checksum. But like a checksum,
1319: you can't derive the original message from its message digest.
1320:
1321: A message digest alone is not enough to authenticate a message. The
1322: message digest algorithm is publicly known, and does not require
1323: knowledge of any secret keys to calculate. If all we did was attach
1324: a message digest to a message, then a forger could alter a message
1325: and simply attach a new message digest calculated from the new
1326: altered message. To provide real authentication, the sender has to
1327: encrypt (sign) the message digest with his secret key.
1328:
1329: A message digest is calculated from the message by the sender. The
1330: sender's secret key is used to encrypt the message digest and an
1331: electronic timestamp, forming a digital signature, or signature
1332: certificate. The sender sends the digital signature along with the
1333: message. The receiver receives the message and the digital
1334: signature, and recovers the original message digest from the digital
1335: signature by decrypting it with the sender's public key. The
1336: receiver computes a new message digest from the message, and checks
1337: to see if it matches the one recovered from the digital signature. If
1338: it matches, then that proves the message was not altered, and it came
1339: from the sender who owns the public key used to check the signature.
1340:
1341: A potential forger would have to either produce an altered message
1342: that produces an identical message digest (which is infeasible), or
1343: he would have to create a new digital signature from a different
1344: message digest (also infeasible, without knowing the true sender's
1345: secret key).
1346:
1347: Digital signatures prove who sent the message, and that the message
1348: was not altered either by error or design. It also provides
1349: non-repudiation, which means the sender cannot easily disavow his
1350: signature on the message.
1351:
1352: Using message digests to form digital signatures has other advantages
1353: besides being faster than directly signing the entire actual message
1354: with the secret key. Using message digests allows signatures to be
1355: of a standard small fixed size, regardless of the size of the actual
1356: message. It also allows the software to check the message integrity
1357: automatically, in a manner similar to using checksums. And it allows
1358: signatures to be stored separately from messages, perhaps even in a
1359: public archive, without revealing sensitive information about the
1360: actual messages, because no one can derive any message content from a
1361: message digest.
1362:
1363: The message digest algorithm used here is the MD5 Message Digest
1364: Algorithm, placed in the public domain by RSA Data Security, Inc.
1365: MD5's designer, Ronald Rivest, writes this about MD5:
1366:
1367: "It is conjectured that the difficulty of coming up with two messages
1368: having the same message digest is on the order of 2^64 operations,
1369: and that the difficulty of coming up with any message having a given
1370: message digest is on the order of 2^128 operations. The MD5
1371: algorithm has been carefully scrutinized for weaknesses. It is,
1372: however, a relatively new algorithm and further security analysis is
1373: of course justified, as is the case with any new proposal of this
1374: sort. The level of security provided by MD5 should be sufficient for
1375: implementing very high security hybrid digital signature schemes
1376: based on MD5 and the RSA public-key cryptosystem."
1377:
1378:
1379:
1380: Compatibility with Previous Versions of PGP
1381: ===========================================
1382:
1383: I'm sorry, PGP version 2.0 is not compatible with PGP version 1.0.
1384: If you have keys generated with version 1.0, you will have to
1385: generate new keys to use with this version. This version of PGP uses
1386: all new algorithms for conventional cryptography, compression, and
1387: message digests, as well as using a much better approach to key
1388: management. There were just too many changes to make it compatible
1389: with the old format messages, signatures, and keys. Perhaps we could
1390: have provided a special conversion utility to convert old keys into
1391: new keys, but we were all tired and wanted to get the new release out
1392: the door. Besides, converting the old keys into new keys would
1393: probably create more problems than it would solve, because we have
1394: changed to a new recommended uniform style for the user ID that
1395: includes the full name and E-mail address in a particular syntax.
1396:
1397: There is compatibility from version 2.0 to higher versions. Because
1398: new features are added, older versions may not always be able to
1399: handle some files created with newer versions.
1400:
1401: We made some effort to design the internal data structures of this
1402: version of PGP to be adaptable to future changes, so that hopefully
1403: you will not be required to discard and regenerate your keys in future
1404: versions.
1405:
1406:
1407: Vulnerabilities
1408: ===============
1409:
1410: No data security system is impenetrable. PGP can be circumvented in
1411: a variety of ways. In any data security system, you have to ask
1412: yourself if the information you are trying to protect is more
1413: valuable to your attacker than the cost of the attack. This should
1414: lead you to protecting yourself from the cheapest attacks, while not
1415: worrying about the more expensive attacks.
1416:
1417: Some of the discussion that follows may seem unduly paranoid, but
1418: such an attitude is appropriate for a reasonable discussion of
1419: vulnerability issues.
1420:
1421:
1422: Compromised Pass Phrase and Secret Key
1423: --------------------------------------
1424:
1425: Probably the simplest attack is if you leave your pass phrase for
1426: your secret key written down somewhere. If someone gets it and also
1427: gets your secret key file, they can read your messages and make
1428: signatures in your name.
1429:
1430: Don't use obvious passwords that can be easily guessed, such as the
1431: names of your kids or spouse. If you make your pass phrase a single
1432: word, it can be easily guessed by having a computer try all the words
1433: in the dictionary until it finds your password. That's why a pass
1434: phrase is so much better than a password. A more sophisticated
1435: attacker may have his computer scan a book of famous quotations to
1436: find your pass phrase. An easy to remember but hard to guess pass
1437: phrase can be easily constructed by some creatively nonsensical
1438: sayings or very obscure literary quotes.
1439:
1440: For further details, see the section "How to Protect Secret Keys from
1441: Disclosure" in the Essential Topics volume of the PGP User's Guide.
1442:
1443:
1444: Public Key Tampering
1445: --------------------
1446:
1447: A major vulnerability exists if public keys are tampered with. This
1448: may be the most crucially important vulnerability of a public key
1449: cryptosystem, in part because most novices don't immediately
1450: recognize it. The importance of this vulnerability, and appropriate
1451: hygienic countermeasures, are detailed in the section "How to Protect
1452: Public Keys from Tampering" in the Essential Topics volume.
1453:
1454: To summarize: When you use someone's public key, make certain it has
1455: not been tampered with. A new public key from someone else should be
1456: trusted only if you got it directly from its owner, or if it has been
1457: signed by someone you trust. Make sure no one else can tamper with
1458: your own public key ring. Maintain physical control of both your
1459: public key ring and your secret key ring, preferably on your own
1460: personal computer rather than on a remote timesharing system. Keep a
1461: backup copy of both key rings.
1462:
1463:
1464: "Not Quite Deleted" Files
1465: -------------------------
1466:
1467: Another potential security problem is caused by how most operating
1468: systems delete files. When you encrypt a file and then delete the
1469: original plaintext file, the operating system doesn't actually
1470: physically erase the data. It merely marks those disk blocks as
1471: deleted, allowing the space to be reused later. It's sort of like
1472: discarding sensitive paper documents in the paper recycling bin
1473: instead of the paper shredder. The disk blocks still contain the
1474: original sensitive data you wanted to erase, and will probably
1475: eventually be overwritten by new data at some point in the future.
1476: If an attacker reads these deleted disk blocks soon after they have
1477: been deallocated, he could recover your plaintext.
1478:
1479: In fact this could even happen accidentally, if for some reason
1480: something went wrong with the disk and some files were accidentally
1481: deleted or corrupted. A disk recovery program may be run to recover
1482: the damaged files, but this often means some previously deleted files
1483: are resurrected along with everything else. Your confidential files
1484: that you thought were gone forever could then reappear and be
1485: inspected by whomever is attempting to recover your damaged disk.
1486: Even while you are creating the original message with a word
1487: processor or text editor, the editor may be creating multiple
1488: temporary copies of your text on the disk, just because of its
1489: internal workings. These temporary copies of your text are deleted
1490: by the word processor when it's done, but these sensitive fragments
1491: are still on your disk somewhere.
1492:
1493: Let me tell you a true horror story. I had a friend, married with
1494: young children, who once had a brief and not very serious affair.
1495: She wrote a letter to her lover on her word processor, and deleted
1496: the letter after she sent it. Later, after the affair was over, the
1497: floppy disk got damaged somehow and she had to recover it because it
1498: contained other important documents. She asked her husband to
1499: salvage the disk, which seemed perfectly safe because she knew she
1500: had deleted the incriminating letter. Her husband ran a commercial
1501: disk recovery software package to salvage the files. It recovered
1502: the files alright, including the deleted letter. He read it, which
1503: set off a tragic chain of events.
1504:
1505: The only way to prevent the plaintext from reappearing is to somehow
1506: cause the deleted plaintext files to be overwritten. Unless you know
1507: for sure that all the deleted disk blocks will soon be reused, you
1508: must take positive steps to overwrite the plaintext file, and also
1509: any fragments of it on the disk left by your word processor. You can
1510: overwrite the original plaintext file after encryption by using the
1511: PGP -w (wipe) option. You can take care of any fragments of the
1512: plaintext left on the disk by using any of the disk utilities
1513: available that can overwrite all of the unused blocks on a disk. For
1514: example, the Norton Utilities for MSDOS can do this.
1515:
1516: Even if you overwrite the plaintext data on the disk, it may still be
1517: possible for a resourceful and determined attacker to recover the
1518: data. Faint magnetic traces of the original data remain on the disk
1519: after it has been overwritten. Special sophisticated disk recovery
1520: hardware can sometimes be used to recover the data.
1521:
1522:
1523: Viruses and Trojan Horses
1524: -------------------------
1525:
1526: Another attack could involve a specially-tailored hostile computer
1527: virus or worm that might infect PGP or your operating system. This
1528: hypothetical virus could be designed to capture your pass phrase or
1529: secret key or deciphered messages, and covertly write the captured
1530: information to a file or send it through a network to the virus's
1531: owner. Or it might alter PGP's behavior so that signatures are not
1532: properly checked. This attack is cheaper than cryptanalytic attacks.
1533:
1534: Defending against this falls under the category of defending against
1535: viral infection generally. There are some moderately capable
1536: anti-viral products commercially available, and there are hygienic
1537: procedures to follow that can greatly reduce the chances of viral
1538: infection. A complete treatment of anti-viral and anti-worm
1539: countermeasures is beyond the scope of this document. PGP has no
1540: defenses against viruses, and assumes your own personal computer is a
1541: trustworthy execution environment. If such a virus or worm actually
1542: appeared, hopefully word would soon get around warning everyone.
1543:
1544: Another similar attack involves someone creating a clever imitation
1545: of PGP that behaves like PGP in most respects, but doesn't work the
1546: way it's supposed to. For example, it might be deliberately crippled
1547: to not check signatures properly, allowing bogus key certificates to
1548: be accepted. This "Trojan horse" version of PGP is not hard for an
1549: attacker to create, because PGP source code is widely available, so
1550: anyone could modify the source code and produce a lobotomized zombie
1551: imitation PGP that looks real but does the bidding of its diabolical
1552: master. This Trojan horse version of PGP could then be widely
1553: circulated, claiming to be from me. How insidious.
1554:
1555: You should make an effort to get your copy of PGP from a reliable
1556: source, whatever that means. Or perhaps from more than one
1557: independent source, and compare them with a file comparison utility.
1558:
1559: There are other ways to check PGP for tampering, using digital
1560: signatures. If someone you trust signs the executable version of
1561: PGP, vouching for the fact that it has not been infected or tampered
1562: with, you can be reasonably sure that you have a good copy. You
1563: could use an earlier trusted version of PGP to check the signature on
1564: a later suspect version of PGP. But this will not help at all if
1565: your operating system is infected, nor will it detect if your
1566: original copy of PGP.EXE has been maliciously altered in such a way
1567: as to compromise its own ability to check signatures. This test also
1568: assumes that you have a good trusted copy of the public key that you
1569: use to check the signature on the PGP executable.
1570:
1571:
1572: Physical Security Breach
1573: ------------------------
1574:
1575: A physical security breach may allow someone to physically acquire
1576: your plaintext files or printed messages. A determined opponent
1577: might accomplish this through burglary, trash-picking, unreasonable
1578: search and seizure, or bribery, blackmail or infiltration of your
1579: staff. Some of these attacks may be especially feasible against
1580: grassroots political organizations that depend on a largely volunteer
1581: staff. It has been widely reported in the press that the FBI's
1582: COINTELPRO program used burglary, infiltration, and illegal bugging
1583: against antiwar and civil rights groups. And look what happened at
1584: the Watergate Hotel.
1585:
1586: Don't be lulled into a false sense of security just because you have
1587: a cryptographic tool. Cryptographic techniques protect data only
1588: while it's encrypted-- direct physical security violations can still
1589: compromise plaintext data or written or spoken information.
1590:
1591: This kind of attack is cheaper than cryptanalytic attacks on PGP.
1592:
1593:
1594: Tempest Attacks
1595: ---------------
1596:
1597: Another kind of attack that has been used by well-equipped opponents
1598: involves the remote detection of the electromagnetic signals from
1599: your computer. This expensive and somewhat labor-intensive attack is
1600: probably still cheaper than direct cryptanalytic attacks. An
1601: appropriately instrumented van can park near your office and remotely
1602: pick up all of your keystrokes and messages displayed on your
1603: computer video screen. This would compromise all of your passwords,
1604: messages, etc. This attack can be thwarted by properly shielding all
1605: of your computer equipment and network cabling so that it does not
1606: emit these signals. This shielding technology is known as "Tempest",
1607: and is used by some Government agencies and defense contractors.
1608: There are hardware vendors who supply Tempest shielding commercially,
1609: although it may be subject to some kind of Government licensing. Now
1610: why do you suppose the Government would restrict access to Tempest
1611: shielding?
1612:
1613:
1614: Exposure on Multi-user Systems
1615: ------------------------------
1616:
1617: PGP was originally designed for a single-user MSDOS machine under
1618: your direct physical control. I run PGP at home on my own PC, and
1619: unless someone breaks into my house or monitors my electromagnetic
1620: emissions, they probably can't see my plaintext files or secret keys.
1621:
1622: But now PGP also runs on multi-user systems such as Unix and VAX/VMS.
1623: On multi-user systems, there are much greater risks of your plaintext
1624: or keys or passwords being exposed. The Unix system administrator or
1625: a clever intruder can read your plaintext files, or perhaps even use
1626: special software to covertly monitor your keystrokes or read what's
1627: on your screen. On a Unix system, any other user can read your
1628: environment information remotely by simply using the Unix "ps"
1629: command. Similar problems exist for MSDOS machines connected on a
1630: local area network. The actual security risk is dependent on your
1631: particular situation. Some multi-user systems may be safe because
1632: all the users are trusted, or because they have system security
1633: measures that are safe enough to withstand the attacks available to
1634: the intruders, or because there just aren't any sufficiently
1635: interested intruders. Some Unix systems are safe because they are
1636: only used by one user-- there are even some notebook computers
1637: running Unix. It would be unreasonable to simply exclude PGP from
1638: running on all Unix systems.
1639:
1640: PGP is not designed to protect your data while it is in plaintext
1641: form on a compromised system. Nor can it prevent an intruder from
1642: using sophisticated measures to read your secret key while it is
1643: being used. You will just have to recognize these risks on
1644: multi-user systems, and adjust your expectations and behavior
1645: accordingly. Perhaps your situation is such that you should consider
1646: only running PGP on an isolated single-user system under your direct
1647: physical control. That's what I do, and that's what I recommend.
1648:
1649:
1650: Traffic Analysis
1651: ----------------
1652:
1653: Even if the attacker cannot read the contents of your encrypted
1654: messages, he may be able to infer at least some useful information by
1655: observing where the messages come from and where they are going, the
1656: size of the messages, and the time of day the messages are sent.
1657: This is analogous to the attacker looking at your long distance phone
1658: bill to see who you called and when and for how long, even though the
1659: actual content of your calls is unknown to the attacker. This is
1660: called traffic analysis. PGP alone does not protect against traffic
1661: analysis. Solving this problem would require specialized
1662: communication protocols designed to reduce exposure to traffic
1663: analysis in your communication environment, possibly with some
1664: cryptographic assistance.
1665:
1666:
1667: Cryptanalysis
1668: -------------
1669:
1670: An expensive and formidable cryptanalytic attack could possibly be
1671: mounted by someone with vast supercomputer resources, such as a
1672: Government intelligence agency. They might crack your RSA key by
1673: using some new secret factoring breakthrough. Perhaps so, but it is
1674: noteworthy that the US Government trusts the RSA algorithm enough in
1675: some cases to use it to protect its own nuclear weapons, according to
1676: Ron Rivest. And civilian academia has been intensively attacking it
1677: without success since 1978.
1678:
1679: Perhaps the Government has some classified methods of cracking the
1680: IDEA(tm) conventional encryption algorithm used in PGP. This is
1681: every cryptographer's worst nightmare. There can be no absolute
1682: security guarantees in practical cryptographic implementations.
1683:
1684: Still, some optimism seems justified. The IDEA algorithm's designers
1685: are among the best cryptographers in Europe. It has had extensive
1686: security analysis and peer review from some of the best cryptanalysts
1687: in the unclassified world. It appears to have some design advantages
1688: over the DES in withstanding differential cryptanalysis, which has
1689: been used to crack the DES.
1690:
1691: Besides, even if this algorithm has some subtle unknown weaknesses,
1692: PGP compresses the plaintext before encryption, which should greatly
1693: reduce those weaknesses. The computational workload to crack it is
1694: likely to be much more expensive than the value of the message.
1695:
1696: If your situation justifies worrying about very formidable attacks of
1697: this caliber, then perhaps you should contact a data security
1698: consultant for some customized data security approaches tailored to
1699: your special needs. Boulder Software Engineering, whose address and
1700: phone are given at the end of this document, can provide such
1701: services.
1702:
1703:
1704: In summary, without good cryptographic protection of your data
1705: communications, it may have been practically effortless and perhaps
1706: even routine for an opponent to intercept your messages, especially
1707: those sent through a modem or E-mail system. If you use PGP and
1708: follow reasonable precautions, the attacker will have to expend far
1709: more effort and expense to violate your privacy.
1710:
1711: If you protect yourself against the simplest attacks, and you feel
1712: confident that your privacy is not going to be violated by a
1713: determined and highly resourceful attacker, then you'll probably be
1714: safe using PGP. PGP gives you Pretty Good Privacy.
1715:
1716:
1717: Legal Issues
1718: ============
1719:
1720:
1721: Trademarks, Copyrights, and Warranties
1722: --------------------------------------
1723:
1724: "Pretty Good Privacy", "Phil's Pretty Good Software", and the "Pretty
1725: Good" label for computer software and hardware products are all
1726: trademarks of Philip Zimmermann and Phil's Pretty Good Software. PGP
1727: is (c) Copyright Philip R. Zimmermann, 1990-1993. Philip Zimmermann
1728: also holds the copyright for the PGP User's Manual, as well as any
1729: foreign language translations of the manual or the software.
1730:
1731: The author assumes no liability for damages resulting from the use of
1732: this software, even if the damage results from defects in this
1733: software, and makes no representations concerning the merchantability
1734: of this software or its suitability for any specific purpose. It is
1735: provided "as is" without express or implied warranty of any kind.
1736:
1737:
1738: Patent Rights on the Algorithms
1739: -------------------------------
1740:
1741: When I first released PGP, I half-expected to encounter some form of
1742: legal harassment from the Government. Indeed, there has been legal
1743: harrassment, but it hasn't come from the Government-- it has come
1744: from a private corporation.
1745:
1746: The RSA public key cryptosystem was developed at MIT with Federal
1747: funding from grants from the National Science Foundation and the
1748: Navy. It is patented by MIT (U.S. patent #4,405,829, issued 20 Sep
1749: 1983). A company in California called Public Key Partners (PKP) holds
1750: the exclusive commercial license to sell and sub-license the RSA
1751: public key cryptosystem. The author of this software implementation
1752: of the RSA algorithm is providing this implementation for educational
1753: use only. Licensing this algorithm from PKP is the responsibility of
1754: you, the user, not Philip Zimmermann, the author of this software
1755: implementation. The author assumes no liability for any patent
1756: infringement that may result from the unlicensed use by the user of
1757: the underlying RSA algorithm used in this software. Foreign users
1758: should note that the RSA patent does not apply outside the US, and
1759: there is no RSA patent in any other country. Federal agencies may
1760: use it because the Government paid for the development of RSA.
1761:
1762: Unfortunately, PKP is not offering any licensing of their RSA patent
1763: to end users of PGP. This essentially makes PGP contraband in the
1764: USA. Jim Bidzos, president of PKP, threatened to take legal action
1765: against me unless I stop distributing PGP, until they can devise a
1766: licensing scheme for it. I agreed to this, since PGP is already in
1767: wide circulation and waiting a while for a licensing arrangement from
1768: PKP seemed reasonable. Mr. Bidzos assured me (he even used the word
1769: "promise") several times since the initial 5 June 91 release of PGP
1770: that they were working on a licensing scheme for PGP. Apparently, my
1771: release of PGP helped provide the impetus for them to offer some sort
1772: of a freeware-style license for noncommercial use of the RSA
1773: algorithm. However, in December 1991 Mr. Bidzos said he had no plans
1774: to ever license the RSA algorithm to PGP users, and denied ever
1775: implying that he would. Meanwhile, I have continued to refrain from
1776: distributing PGP, although I continue to update the PGP User's Guide,
1777: and have provided the design guidance for new revisions of PGP.
1778: Ironically, all this legal controversy from PKP has imparted a
1779: forbidden flavor to PGP that has only served to amplify its universal
1780: popularity.
1781:
1782: I wrote my PGP software from scratch, with my own implementation of
1783: the RSA algorithm. I didn't steal any software from PKP. Before
1784: publishing PGP, I got a formal written legal opinion from a patent
1785: attorney with extensive experience in software patents. I'm
1786: convinced that publishing PGP the way I did does not violate patent
1787: law. However, it is a well known axiom in the US legal system that
1788: regardless of the law, he with the most money and lawyers prevails,
1789: if not by actually winning then by crushing the little guy with legal
1790: expenses.
1791:
1792: Not only did PKP acquire the exclusive patent rights for the RSA
1793: cryptosystem, which was developed with your tax dollars, but they
1794: also somehow acquired the exclusive rights to three other patents
1795: covering rival public key schemes invented by others, also developed
1796: with your tax dollars. This essentially gives one company a legal
1797: lock in the USA on nearly all practical public key cryptosystems.
1798: They even appear to be claiming patent rights on the very concept of
1799: public key cryptography, regardless of what clever new original
1800: algorithms are independently invented by others. And you thought
1801: patent law was designed to encourage innovation! PKP does not
1802: actually develop any software-- they don't even have an engineering
1803: department-- they are essentially a litigation company.
1804:
1805: Public key cryptography is destined to become a crucial technology in
1806: the protection of our civil liberties and privacy in our increasingly
1807: connected society. Why should the Government try to limit access to
1808: this key technology, when a single monopoly can do it for them?
1809:
1810: It appears certain that there will be future releases of PGP,
1811: regardless of the outcome of licensing problems with Public Key
1812: Partners. If PKP does not license PGP, then future releases of PGP
1813: might not come from me. There are countless fans of PGP outside the
1814: US, and many of them are software engineers who want to improve PGP
1815: and promote it, regardless of what I do. The second release of PGP
1816: was a joint effort of an international team of software engineers,
1817: implementing enhancements to the original PGP with design guidance
1818: from me. It was released by Branko Lankester in The Netherlands and
1819: Peter Gutmann in New Zealand, out of reach of US patent law.
1820: Although released only in Europe and New Zealand, it spontaneously
1821: spread to the USA without help from me or the PGP development team.
1822:
1823: The IDEA(tm) conventional block cipher used by PGP is covered by a
1824: patent in Europe, held by ETH and a Swiss company called Ascom-Tech
1825: AG. The patent number is PCT/CH91/00117. International patents are
1826: pending. IDEA(tm) is a trademark of Ascom-Tech AG. There is no
1827: license fee required for noncommercial use of IDEA. Ascom Tech AG
1828: has granted permission for PGP to use the IDEA cipher, and places no
1829: restrictions on using PGP for any purpose, including commercial use.
1830: You may not extract the IDEA cipher from PGP and put it in another
1831: commercial product without a license. Commercial users of IDEA may
1832: obtain licensing details from Dieter Profos, Ascom Tech AG, Solothurn
1833: Lab, Postfach 151, 4502 Solothurn, Switzerland, Tel +41 65 242885,
1834: Fax +41 65 235761.
1835:
1836: The ZIP compression routines in PGP come from freeware source code,
1837: with the author's permission. I'm not aware of any patents on the
1838: ZIP algorithm, but you're welcome to check into that question
1839: yourself. If there are any obscure patent claims that apply to ZIP,
1840: then sorry, you'll have to take care of the patent licensing, not me.
1841:
1842: All this patent stuff reminds me of a Peanuts cartoon I saw in the
1843: newspaper where Lucy showed Charlie Brown a fallen autumn leaf and
1844: said "This is the first leaf to fall this year." Charlie Brown said,
1845: "How do you know that? Leaves have been falling for weeks." Lucy
1846: replied, "I had this one notarized."
1847:
1848:
1849: Licensing and Distribution
1850: --------------------------
1851:
1852: In the USA PKP controls, through US patent law, the licensing of the
1853: RSA algorithm. But I have no objection to anyone freely using or
1854: distributing my PGP software, without payment of fees to me. You must
1855: keep the copyright notices on PGP and keep this documentation with
1856: it. However, if you live in the USA, this may not satisfy any legal
1857: obligations you may have to PKP for using the RSA algorithm as
1858: mentioned above.
1859:
1860: In fact, if you live in the USA, and you are not a Federal agency,
1861: you shouldn't actually run PGP on your computer, because Public Key
1862: Partners wants to forbid you from running my software. PGP is
1863: contraband.
1864:
1865: Of course, I can't give any assurances, but my guess is that it seems
1866: unlikely that PKP would waste their time pursuing PGP end users for
1867: patent infringement. There are just too many PGP users to go after.
1868: And why would they single you out? But I certainly wouldn't want to
1869: imply that you do anything improper-- if PKP were offering licenses,
1870: I would urge you to obtain one. But since they aren't, well, I guess
1871: you should just refrain from using PGP if you live in the USA.
1872:
1873: PGP is not shareware, it's freeware. Forbidden freeware. Published
1874: as a community service. If I sold PGP for money, then I would get
1875: sued by PKP for using their RSA algorithm. More importantly, giving
1876: PGP away for free will encourage far more people to use it, which
1877: hopefully will have a greater social impact. This could lead to
1878: widespread awareness and use of the RSA public key cryptosystem,
1879: which will probably make more money for PKP in the long run. If only
1880: they could see that.
1881:
1882: All the source code for PGP is available for free under the "Copyleft"
1883: General Public License from the Free Software Foundation (FSF). A
1884: copy of the FSF General Public License is included in the source
1885: release package of PGP.
1886:
1887: Regardless of and perhaps contrary to some provisions of the FSF
1888: General Public License, the following terms apply:
1889:
1890: 1) Written discussions of PGP in magazines or books may include
1891: fragments of PGP source code and documentation, without
1892: restrictions.
1893:
1894: 2) Although the FSF General Public License allows non-proprietary
1895: derivative products, it prohibits proprietary derivative products.
1896: Despite this, I may grant you a special license if you want to
1897: derive a proprietary commercial product from some of PGP's parts.
1898: There may or may not be a fee depending on what kind of a deal you
1899: plan to pursue with PKP. Retaining my copyright notice and
1900: attribution might suffice in some cases. Give me a call and we'll
1901: discuss it. I'm real easy to please.
1902:
1903: Feel free to disseminate the complete PGP release package as widely
1904: as possible. Give it to all your friends. If you have access to any
1905: electronic Bulletin Boards Systems, please upload the complete PGP
1906: executable object release package to as many BBS's as possible. You
1907: may disseminate the PGP source release package too, if you've got
1908: it. The PGP version 2.2 executable object release package for MSDOS
1909: contains the PGP executable software, documentation, sample key rings
1910: including my own public key, and signatures for the software and this
1911: manual, all in one PKZIP compressed file called PGP22.ZIP. The PGP
1912: source release package for MSDOS contains all the C source files in
1913: one PKZIP compressed file called PGP22SRC.ZIP.
1914:
1915: You may obtain free copies or updates to PGP from thousands of BBS's
1916: worldwide or from other public sources such as Internet FTP sites.
1917: Don't ask me for a copy directly from me, since I'd rather avoid
1918: further legal problems with PKP at this time. I might be able to
1919: tell you where you can get it, however.
1920:
1921: After all this work I have to admit I wouldn't mind getting some fan
1922: mail for PGP, to gauge its popularity. Let me know what you think
1923: about it and how many of your friends use it. Bug reports and
1924: suggestions for enhancing PGP are welcome, too. Perhaps a future PGP
1925: release will reflect your suggestions.
1926:
1927: This project has not been funded and the project has nearly eaten me
1928: alive. This means you can't count on a reply to your mail, unless
1929: you only need a short written reply and you include a stamped
1930: self-addressed envelope. But I do reply to E-mail. Please keep it in
1931: English, as my foreign language skills are weak. If you call and I'm
1932: not in, it's best to just try again later. I usually don't return
1933: long distance phone calls, unless you leave a message that I can call
1934: you collect. If you need any significant amount of my time, I am
1935: available on a paid consulting basis, and I do return those calls.
1936:
1937: The most inconvenient mail I get is for some well-intentioned person
1938: to send me a few dollars asking me for a copy of PGP. I can't send
1939: it to them because of the legal threats from PKP (or worse--
1940: sometimes these requests are from foreign countries, and I would be
1941: risking violating cryptographic export control laws). Even if there
1942: were no legal hassles involved in sending PGP to them, they usually
1943: don't send enough money to make it worth my time ($50 might be worth
1944: my time if I were actually selling this stuff). I'm just not set up
1945: as a low cost low volume mail order business. I can't just ignore
1946: the request and keep the money, because they probably regard the
1947: money as a fee for me to fulfill their request. If I return the
1948: money, I might have to get in my car and drive down to the post
1949: office and buy some postage stamps, because these requests rarely
1950: include a stamped self-addressed envelope. And I have to take the
1951: time to write a polite reply that I can't do it. If I postpone the
1952: reply and set the letter down on my desk, it might be buried within
1953: minutes and won't see the light of day again for months. Multiply
1954: these minor inconveniences by the number of requests I get, and you
1955: can see the problem. Isn't it enough that the software is free? It
1956: would be nicer if people could try to get PGP from any of the myriad
1957: other sources. If you don't have a modem, ask a friend to get it for
1958: you. If you can't find it yourself, I don't mind answering a quick
1959: phone call.
1960:
1961: If anyone wants to volunteer to improve PGP, please let me know. It
1962: could certainly use some more work. Some features were deferred to
1963: get it out the door. A number of PGP users have since donated their
1964: time to port PGP to Unix on Sun SPARCstations, to Ultrix, to VAX/VMS,
1965: to OS/2, to the Amiga, and to the Atari ST. Perhaps you can help
1966: port it to some new environments. But please let me know if you plan
1967: to port or add enhancements to PGP, to avoid duplication of effort,
1968: and to avoid starting with an obsolete version of the source code.
1969:
1970: Because so many foreign language translations of PGP have been
1971: produced, most of them are not distributed with the regular PGP
1972: release package because it would require too much disk space.
1973: Separate language translation "kits" are available from a number of
1974: independent sources, and are sometimes available separately from the
1975: same distribution centers that carry the regular PGP release
1976: software. These kits include translated versions of the file
1977: LANGUAGE.TXT, PGP.HLP, and the PGP User's Guide. If you want to
1978: produce a translation for your own native language, contact me first
1979: to get the latest information and standard guidelines, and to find
1980: out if it's been translated to your language already. Colin Plumb
1981: ([email protected]) maintains a comprehensive collection of foreign
1982: language kits from other translators.
1983:
1984: Future versions of PGP may have to change the data formats for
1985: messages, signatures, keys and key rings, in order to provide
1986: important new features. This may cause backward compatibility
1987: problems with this version of PGP. Future releases may provide
1988: conversion utilities to convert old keys, but you may have to dispose
1989: of old messages created with the old PGP.
1990:
1991: If you have access to the Internet, watch for announcements of new
1992: releases of PGP on the Internet newsgroups "sci.crypt" and PGP's own
1993: newsgroup, "alt.security.pgp". There is also an interest group
1994: mailing list called info-pgp, which is intended for users without
1995: access to the "alt.security.pgp" newsgroup. Info-pgp is moderated by
1996: Hugh Miller, and you may subscribe to it by writing him a letter at
1997: [email protected]. Include your name and Internet
1998: address. If you want to know where to get PGP, Hugh can send you a
1999: list of Internet FTP sites and BBS phone numbers. Hugh may also be
2000: reached at [email protected].
2001:
2002:
2003:
2004: Export Controls
2005: ---------------
2006:
2007: The Government has made it illegal in many cases to export good
2008: cryptographic technology, and that may include PGP. They regard this
2009: kind of software as munitions. This is determined by volatile State
2010: Department policies, not fixed laws. I will not export this software
2011: out of the US or Canada in cases when it is illegal to do so under US
2012: State Department policies, and I assume no responsibility for other
2013: people exporting it on their own.
2014:
2015: If you live outside the US or Canada, I advise you not to violate US
2016: State Department policies by getting PGP from a US source. Since
2017: thousands of domestic users got it after its initial publication, it
2018: somehow leaked out of the US and spread itself widely abroad, like
2019: dandelion seeds blowing in the wind. If PGP has already found its
2020: way into your country, then I don't think you're violating US export
2021: law if you pick it up from a source outside of the US.
2022:
2023: It seems to some legal observers I've talked with, that the framers of
2024: the US export controls never envisioned that they would ever apply to
2025: cryptographic freeware that has been published and scattered to the
2026: winds. It's hard to imagine a US attorney trying to build a real
2027: case against someone for the "export" of software published freely in
2028: the US. As far as anyone I've talked to knows, it's never been done,
2029: despite numerous examples of export violations. I'm not a lawyer and
2030: I'm not giving you legal advice-- I'm just trying to point out what
2031: seems like common sense.
2032:
2033: Starting with PGP version 2.0, the release point of the software has
2034: been outside the US, on publicly-accessible computers in Europe.
2035: Each release is electronically sent back into the US and posted on
2036: publicly-accessible computers in the US by PGP privacy activists in
2037: foreign countries. There are some restrictions in the US regarding
2038: the import of munitions, but I'm not aware of any cases where this
2039: was ever enforced for importing cryptographic software into the US.
2040: I imagine that a legal action of that type would be quite a spectacle
2041: of controversy.
2042:
2043: Some foreign governments impose serious penalties on anyone inside
2044: their country for merely using encrypted communications. In some
2045: countries they might even shoot you for that. But if you live in
2046: that kind of country, perhaps you need PGP even more.
2047:
2048:
2049:
2050: Computer-Related Political Groups
2051: =================================
2052:
2053: PGP is a very political piece of software. It seems appropriate to
2054: mention here some computer-related activist groups. Full details on
2055: these groups, and how to join them, is provided in a separate
2056: document file in the PGP release package.
2057:
2058: The Electronic Frontier Foundation (EFF) was founded in July, 1990,
2059: to assure freedom of expression in digital media, with a particular
2060: emphasis on applying the principles embodied in the Constitution and
2061: the Bill of Rights to computer-based communication. They can be
2062: reached at: Electronic Frontier Foundation, 238 Main Street,
2063: Cambridge, MA 02142, USA.
2064:
2065: The League for Programming Freedom (LPF) is a grass-roots organization
2066: of professors, students, businessmen, programmers and users dedicated
2067: to bringing back the freedom to write programs. They regard patents
2068: on computer algorithms as harmful to the US software industry. They
2069: can be reached at (617) 433-7071, or send Internet mail to
2070: [email protected]
2071:
2072: Computer Professionals For Social Responsibility (CPSR) empowers
2073: computer professionals and computer users to advocate for the
2074: responsible use of information technology and empowers all who use
2075: computer technology to participate in public policy debates on the
2076: impacts of computers on society. They can be reached at:
2077: 415-322-3778 in Palo Alto, E-mail address [email protected].
2078:
2079: For more details on these groups, see the accompanying document in
2080: the PGP release package.
2081:
2082:
2083: Recommended Introductory Readings
2084: =================================
2085:
2086: 1) Dorothy Denning, "Cryptography and Data Security", Addison-Wesley,
2087: Reading, MA 1982
2088: 2) Dorothy Denning, "Protecting Public Keys and Signature Keys",
2089: IEEE Computer, Feb 1983
2090: 3) Martin E. Hellman, "The Mathematics of Public-Key Cryptography,"
2091: Scientific American, Aug 1979
2092:
2093: Other Readings
2094: ==============
2095:
2096: 4) Ronald Rivest, "The MD5 Message Digest Algorithm", MIT Laboratory
2097: for Computer Science, 1991
2098: 5) Xuejia Lai, "On the Design and Security of Block Ciphers",
2099: Institute for Signal and Information Processing, ETH-Zentrum,
2100: Zurich, Switzerland, 1992
2101: 6) Xuejia Lai, James L. Massey, Sean Murphy, "Markov Ciphers and
2102: Differential Cryptanalysis", Advances in Cryptology- EUROCRYPT'91
2103: 7) Philip Zimmermann, "A Proposed Standard Format for RSA
2104: Cryptosystems", Advances in Computer Security, Vol III, edited by
2105: Rein Turn, Artech House, 1988
2106: 8) Paul Wallich, "Electronic Envelopes", Scientific American, Feb
2107: 1993, pages 30-32. (This is an article on PGP)
2108:
2109:
2110: To Contact the Author
2111: =====================
2112:
2113: Philip Zimmermann may be reached at:
2114:
2115: Boulder Software Engineering
2116: 3021 Eleventh Street
2117: Boulder, Colorado 80304 USA
2118: Phone 303-541-0140 (voice or FAX) (10:00am - 7:00pm Mountain Time)
2119: Internet: [email protected]
2120:
2121:
2122:
2123: Appendix A: Where to Get PGP
2124: =============================
2125:
2126: The following describes how to get the freeware public key
2127: cryptographic software PGP (Pretty Good Privacy) from an anonymous
2128: FTP site on Internet, or from other sources.
2129:
2130: PGP has sophisticated key management, an RSA/conventional hybrid
2131: encryption scheme, message digests for digital signatures, data
2132: compression before encryption, and good ergonomic design. PGP is
2133: well featured and fast, and has excellent user documentation. Source
2134: code is free.
2135:
2136: PGP uses the RSA cryptosystem which is claimed by a US patent held by
2137: a company called Public Key Partners. PGP users outside the US take
2138: note that there is no RSA patent outside the US. Also, bear in mind
2139: that there are US and Canadian export laws prohibiting anyone inside
2140: the US and Canada from exporting cryptographic software like this.
2141: If you live outside the US, you're probably not violating US export
2142: law if you pick it up from a source outside of the US.
2143:
2144: What follows is a small sample of places that allegedly have PGP, as
2145: of 2 March 1993. This information is not guaranteed to be correct.
2146: Some US sites have occasionally withdrawn PGP because of fear of
2147: legal intimidation from the RSA patent holders.
2148:
2149: There are two compressed archive files in the PGP 2.2 MSDOS release.
2150: You must get pgp22.zip which contains the MSDOS binary executable and
2151: the PGP User's Guide, and you can optionally get pgp22src.zip which
2152: contains the source files. These files can be decompressed with the
2153: MSDOS shareware archive decompression utility PKUNZIP.EXE, version
2154: 1.1 or later. For Unix users who lack an implementation of UNZIP,
2155: the source code can also be found in the compressed tar file
2156: pgp22src.tar.Z.
2157:
2158: A reminder: Set mode to binary or image when doing an FTP transfer.
2159: And when doing a kermit download to your PC, specify 8-bit binary
2160: mode at both ends. Here are some Internet sites that have PGP via
2161: anonymous FTP:
2162:
2163: Finland: nic.funet.fi (128.214.6.100)
2164: Directory: /pub/unix/security/crypt/
2165:
2166: Italy: ghost.dsi.unimi.it (149.132.2.1)
2167: Directory: /pub/security/
2168:
2169: UK: src.doc.ic.ac.uk
2170: Directory: /computing/security/software/PGP
2171:
2172: For those lacking FTP connectivity to the net, nic.funet.fi also
2173: offers the files via email. Send the following mail message to
2174: [email protected]:
2175:
2176: ENCODER uuencode
2177: SEND pub/unix/security/crypt/pgp22src.zip
2178: SEND pub/unix/security/crypt/pgp22.zip
2179:
2180: This will deposit the two zipfiles, as (about) 15 batched messages in
2181: your mailbox within about 24 hours. Save and uudecode.
2182:
2183: In the US, PGP may be found on God knows how many BBS systems, far
2184: too many to list here. Still, if you don't have any local BBS phone
2185: numbers handy, here are some BBS's you might try.
2186:
2187: The GRAPEVINE BBS in Little Rock Arkansas has set up a special
2188: account for people to download PGP for free. The SYSOP is Jim Wenzel,
2189: at [email protected]. The following phone numbers are
2190: applicable and should be dialed in the order presented (i.e., the
2191: first one is the highest speed line): (501) 753-6859, (501)
2192: 753-8121, (501) 791-0124. When asked to login use the following
2193: information:
2194:
2195: name: PGP USER ('PGP' is 1st name, 'USER' is 2nd name)
2196: password: PGP
2197:
2198: The Northern Lights BBS in Troy, NY, has PGP for free download. It
2199: is run by Daniel Ray. Call (518) 237-2163 at 300-2400 bps 8N1. Then
2200: login directly to the pgp account as follows:
2201:
2202: tnllogin: pgp
2203: Password: key
2204:
2205: In Colorado, try this single-line BBS: 303 443-8292. It is often
2206: busy, so keep trying. Log in with your own name.
2207:
2208: PGP is also widely available on Fidonet, a large informal network of
2209: PC-based bulletin board systems interconnected via modems. Check
2210: your local bulletin board systems. It is available on many foreign
2211: and domestic Fidonet BBS sites.
2212:
2213: In New Zealand, try this (supposedly free) dial-up BBS system:
2214: Kappa Crucis: +64 9 817-3714, -3725, -3324, -8424, -3094, -3393
2215:
2216: For information on PGP implementations on the Apple Macintosh,
2217: Commodore Amiga, or Atari ST, or any other questions about where to
2218: get PGP for any other platform, contact Hugh Miller at
2219: [email protected].
2220:
2221: Here are a few people and their email addresses or phone numbers you
2222: can contact in some countries to get information on local PGP
2223: availability:
2224:
2225: Peter Gutmann Hugh Kennedy
2226: [email protected] [email protected]
2227: New Zealand Germany
2228:
2229: Branko Lankester Miguel Angel Gallardo
2230: [email protected] [email protected]
2231: +31 2159 42242 (341) 474 38 09
2232: The Netherlands Spain
2233:
2234: Hugh Miller Colin Plumb
2235: [email protected] [email protected]
2236: (312) 508-2727 Toronto, Ontario, Canada
2237: USA
2238:
2239: Jean-loup Gailly
2240: [email protected]
2241: France
2242:
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.