Welcome to the GnuPG Frequently Asked Questions (FAQ)! Before we begin, there’s just a few things that need to be addressed. It’s regrettable these things have to be included, but society has become very litigious.

• GnuPG Desktop and GnuPG VS-Desktop are trademarks of g10 Code GmbH.
• GNU is a trademark of the Free Software Foundation.
• Linux is a trademark of Linus Torvalds.
• Macintosh, OS X, macOS, and Mac OS X are all trademarks of the Apple Corporation.
• PGP is a trademark of CA, Inc.
• Solaris is a trademark of Oracle Corporation.
• UNIX is a trademark of The Open Group.
• Windows is a trademark of the Microsoft Corporation.
• Some cryptographic algorithms mentioned in this FAQ may be trademarked.

The use of these, or any other, marks is solely for identification purposes.

This document is © 2012-2018, Robert J. Hansen <rjh@sixdemonbag.org> and A.M. Kuchling <amk@amk.ca>. You are free to make use of this document in accordance with the Creative Commons Attribution-ShareAlike 3.0 license; alternately, you may make use of it under terms of the GNU General Public License (version 3 or, at your discretion, any later version).

Although the contents of this document are believed to be correct, the author(s) cannot promise it is error-free and cannot assume liability for any errors.

Welcome to the official GnuPG FAQ. Like all FAQs, this is a work in progress. If you have questions that you think should be on it but aren’t, please feel free to email the FAQ maintainer (Rob Hansen, rjh@sixdemonbag.org) or bring your suggestion up on GnuPG-Users.

Thanks to the Free Software Foundation, this FAQ is also available in Russian.

First, please don’t send emails directly to people in GnuPG. While we will try to help to people who send email directly to us, those emails quickly accumulate. Helping just six people a day can take an hour of time, and that’s an hour less we have to work on making GnuPG better. Please reach out to the GnuPG community via the GnuPG-Users mailing list, not individual people within GnuPG.

Second, tell us your operating environment. Be as specific as possible. What operating system are you using? Which version of GnuPG are you using? Where did you get GnuPG from? If your problem is related to email, which email client are you using? Which version number? Is GnuPG supported natively, or is there a plugin? If so, what’s the version number of that?

Third, tell us your problem. Be as specific as possible.

Do this, and you might be surprised at how quickly your problem is solved. An example of a good question would be, “I’m running GnuPG 1.4.14 on an Ubuntu 15.04 x64 box. I’m using Thunderbird with Enigmail. Everything was fine until I did a software update. Ever since then I can’t use GnuPG with email. What happened?” This question gives us enough to work with, and in short order someone will have an answer for you.

A bad question would be, “How do I uninstall GnuPG?” We can’t help you at all; you’ve not given us any of the information we need to answer your question.

Rob Hansen. Please feel free to contact me should there be an error in this FAQ, whether typographical, grammatical, or factual.

When writing, the editorial “we” refers to the general consensus of the GnuPG community. This consensus is hammered out on the GnuPG-Users mailing list. All members of the GnuPG community are invited to participate. Individual people within the community may give their own editorial comments: these will be set off by square brackets, italicized, and initialed by their author. The different editors are:

• wk: Werner Koch <wk@gnupg.org>
• rjh: Robert J. Hansen <rjh@sixdemonbag.org>

Yes.

October 2017.

GnuPG is free cryptographic software from the GNU Project which helps people ensure the confidentiality, integrity and assurance of their data. Let’s try that again: GnuPG is…

• Free. When we say “free” we mean liberty, not price. You’re free to use it, modify it, share it, tinker with it, and learn from it. Software is meant to be used by people, and people deserve freedom.
• Cryptographic. The word “cryptography” is derived from two Greek words, κρυπτός (pronounced “kryptos,” meaning “hidden”) and γράφω (pronounced “grapho,” meaning “writing”). Cryptography is the mathematical study of codes and ciphers.
• Software. This one should already be obvious.
• GNU Project. The GNU Project is a group that aims to give people the ability to do all their computing with free software.
• Confidentiality. No one except authorized parties should be able to read your data.
• Integrity. It shouldn’t be possible to tamper with a message unnoticeably.
• Assurance. An assurance is not a guarantee. There are no guarantees in life, and software is no different. An assurance just means there is good reason to be confident of something — here, it means that when GnuPG is correctly used, people may be confident the data is confidential and/or possesses integrity.

GnuPG may be used by itself as a command-line application (i.e., to be run at a Terminal prompt or a Windows command prompt), or integrated into popular email clients. It’s also used by some instant messaging clients, such as Psi.

“GNU”, followed by the letters “P” and “G”. To learn how to pronounce GNU, you may want to listen to a recording.

Largely, yes. It can be made to interoperate with anything from PGP 5 and onwards, and has excellent interoperability with the most recent releases.

Yes. “Diffie-Hellman” is what PGP calls the Elgamal encryption algorithm. If your PGP-generated keypair uses a Diffie-Hellman encryption subkey, it will appear in GnuPG as an Elgamal subkey. The correct name, incidentally, is Elgamal.

Yes. SHA-256 and SHA-512 belong to a group of hashes known collectively as “SHA-2”. PGP calls SHA-256 and SHA-512 by the non-standard names “SHA-2-256” and “SHA-2-512”, but they are the same algorithms.

Too many to list! It’s known to run on Microsoft Windows, Mac OS X, the various free Unixes, AIX, Solaris, HPUX, OpenVMS, and more. People are even working on porting it to smartphones such as Android.

There is no fixed price. Many sites on the internet offer legal downloads of it for free.

Lots of different places, but no one site hosts binaries for all operating systems.

A convenient Windows installer is available from GPG4WIN.

The GPGtools project has everything needed to get started. Another excellent resources is the GnuPG for OS X project on SourceForge. Finally, Homebrew, Fink, and MacPorts all have it in their repositories.

There are as many ways to install it as there are free Unix systems.

A port to VMS is maintained by Steven M. Schweda at antinode.info.

Yes! The person, business or group that provided you with the GnuPG binary is required to give you the source code upon your request.

The word “free” should evoke ideas of liberty, not price. An awful lot of the software industry does not respect your freedoms: your freedom to use the software for any purpose, your freedom to study and learn from how it works, your freedom to share it with others who might benefit from it, and more. Free Software is the antithesis of this: Free Software is meant to respect your freedoms. You may use the software for any purpose: you may inspect and modify the source code: you may share the software and/or your modifications with others.

GnuPG works with many operating systems, including ones that don’t respect your freedoms. If you’d like to make the shift to a free operating system, you have many choices.

The best way is to visit the donation page.

Development discussion takes place on the gnupg-devel mailing list. Go to the GnuPG mailing list page for links to subscribe and to the list’s archives.

The GnuPG project’s bug tracker is also publicly available.

The good news is the internet is a treasure trove of information. The bad news is that the internet is a festering sewer of misinformation, conspiracy theories, and half-informed speculations all masquerading as informed commentary.

The following mailing lists and web pages are generally known for having a strong signal-to-noise ratio. Nevertheless, we strongly urge you to keep a skeptical mind at all times.

Unfortunately, we can’t help you. If you lose your passphrase, you’ll be unable to use that certificate to sign any new documents or decrypt any existing documents. You can still use it to verify signatures, though. (Technically you could encrypt documents, too, but without the passphrase there’s really not much point: how would you ever decrypt them?)

If you can’t remember your passphrase, the best thing to do is use your pre-made revocation certificate to revoke your old certificate, upload the revocation to the keyserver network, and start anew with a fresh certificate.

First, beware of all absolutes. Almost every question in either the fields of computer security or cryptography can honestly be answered with, “it depends.” Real experts will avoid giving blanket yes-or-no answers except to the simplest and most routine of questions. They will instead hem and haw and explain the several different factors that must be weighed. Hucksters will promise you absolute truth.

Second, the experts really don’t care whether you take their advice. Hucksters often want to be seen as authorities, and if you fail to take their advice they may harangue you about how you’re taking chances with your data, how you’re acting irresponsibly, and so on.

Third, experts genuinely don’t want you to trust them. An expert will instead point to the published literature (usually in a dead-tree edition with the imprimatur of a reputable publishing house) and tell you what the reference books say. They want you to trust the reference books, not them. Hucksters will go on about their extensive personal experience or refer to papers that have only ever been self-published on websites.

Fourth, experts try not to scare people. The world is a scary enough place without it being made moreso. Hucksters will try to scare you, in order to keep you listening to them and dependent on them for information on how to be ‘safe.’

Fifth, experts will quickly admit when they are wrong and give credit to the person bringing the error to their attention. Hucksters tend to take challenges as personal affronts.

There are many excellent mailing lists out there. The following is a list of just some of them that we’ve found to be high-quality. There are undoubtedly many more that we’ve missed.

Subscribing

visit the GnuPG-Users webpage

Unsubscribing

see above

List moderator

<gnupg-users-owner@gnupg.org>

Supports PGP/MIME?

Yes

Languages supported

English

GnuPG-Users is home to the largest community of GnuPG users on the net. The list is very lightly moderated and somewhat freewheeling, but overall it has an excellent signal-to-noise ratio. The level of technical discussion is sometimes a little daunting for the newcomer, but on the whole it’s a wonderful resource.

Subscribing

Visit the Enigmail mailing list page

Unsubscribing

See above

List moderator(s)

• John Clizbe <john@enigmail.net>
• Olav Seyfarth <olav@enigmail.net>
• Patrick Brunschwig <patrick@enigmail.net>
• Ludwig Hügelschäfer <ludwig@enigmail.net>
• Daniele Raffo <dan@enigmail.net>
• Robert J. Hansen <rob@enigmail.net>

Supports PGP/MIME

Yes

Languages supported

English, Deutsch, Schwyzerdütsch, Español

Enigmail integrates GnuPG with Mozilla Thunderbird and/or Mozilla Seamonkey. It’s one of the most popular ways to use GnuPG, and the mailing list provides a friendly place to learn how it works and get started using it.

The list is lightly moderated.

Subscribing

visit the PGPNET page

Unsubscribing

see above

List moderator(s)

Paul Kapaldo <pjkapaldo@yahoo.com>

Supports PGP/MIME?

Yes

Languages supported

Unknown

PGPNET exists to provide people with the opportunity to practice sending and receiving encrypted, signed, and encrypted-and-signed traffic in a group environment.

As a general rule, the huckster quotient of webpages at-large is fairly high. That said, there are some web resources we recommend. They can be broken up into homepages for specific GnuPG-related projects, and sites of general interest.

Many of the projects associated with GnuPG maintain their own websites. If you have problems with an associated project, please check their website first: they might be able to give you faster and better help than the GnuPG community can.

Although the GnuPG community generally finds these websites to be useful, your mileage may significantly vary. There are wide differences of opinion about some of them. They’re worth visiting and worth reading, but make sure to read skeptically.

Many email clients offer strong GnuPG integration.

The column “Active” in the tables below indicate whether the software is actively developed.

(1) With the Enigmail plugin, Thunderbird becomes one of the most popular GnuPG-aware email clients. It’s under active development and is compatible with the latest Thunderbird releases, with a friendly and welcoming user community.

(2) Kontact is KDE’s integrated personal information manager of KDE. It runs anywhere that KDE does, and even on some mobile devices as Kontact Touch.

(3) Claws-Mail for Windows is included in the Gpg4win installer.

(1) With the Enigmail plugin, Thunderbird becomes one of the most popular GnuPG-aware email clients. It’s under active development and is compatible with the latest Thunderbird releases, with a friendly and welcoming user community.

(2) EasyPG is part of Emacs 23, proper as EPA and the underlying EPG. Thus there is no more need to install the plugin. See the Gnus manual for configuration hints. Both EPA and EPG can be customized with the customize-group command and using either the epa or epg groups.

(3) For the best experience make sure to put set crypt_use_gpgme in your ~/.muttrc file. Note that this requires installing Mutt or Neomutt compiled with GPGME support, refer to the Mutt or Neomutt documentation for details.

(4) As of this writing, Apple Mail is incompatible with PGP/MIME. This is a known bug and people are working on it.

(1) With the Enigmail plugin, Thunderbird becomes one of the most popular GnuPG-aware email clients. It’s under active development and is compatible with the latest Thunderbird releases, with a friendly and welcoming user community.

(2) EasyPG is part of Emacs 23, proper as EPA and the underlying EPG. Thus there is no more need to install the plugin. See the Gnus manual for configuration hints. Both EPA and EPG can be customized with the customize-group command and using either the epa or epg groups.

(3) For the best experience make sure to put set crypt_use_gpgme in your ~/.muttrc file. Note that this requires installing Mutt or Neomutt compiled with GPGME support, refer to the Mutt or Neomutt documentation for details.

(4) Kontact is KDE’s integrated personal information manager of KDE. It runs anywhere that KDE does, and even on some mobile devices as Kontact Touch.

Yes, but we don’t recommend it. Sharing a USB token between lots of random computers is a great way to get infested with malware, and that’s not something you want to happen to the token you’re using for secure email. If you’re going to do this, please show caution with respect to which computers you use the portable app on.

That said, Windows users should check PortableApps. Or, to build your own, use the mkportable tool which comes with Gpg4win.

Cryptography tends to use a whole lot of specialized language and jargon. In this section some of it will be deciphered.

In the 1970s new ideas came to the forefront of the cryptanalytic world. One of the most important was the development of asymmetric cryptography (also often called “public-key cryptography”).

Asymmetric cryptography is built around problems that are very hard in one direction, and very easy in another. Consider the number 2,701. If you were to be asked for its prime factors, you would find it a daunting challenge. If you were to be given the numbers 37 and 73, though, it wouldn’t take but a minute to discover the answer was 2,701. Multiplying two numbers to yield a third number is easy: finding those two numbers, given the third, is hard.

Asymmetric cryptography uses these asymmetric problems as the building-blocks of cryptography. It’s easy to create an encrypted message which neither you nor anyone else save the intended recipient can decrypt. To continue the metaphor, you and everyone else get to wrestle with the hard problem (“factor 2,701”). The intended recipient knows a secret piece of information which makes the problem easy (“factor 2,701, given that one of the factors is 73”).

This manages to overcome the major flaw with symmetric cryptography. Your public key can be shared with the entire world, even your enemies, and your communications will still be secure. Compare this to symmetric cryptography, where as soon as the key became public knowledge the entire system was broken.

One of the earliest ciphers was the shift cipher, which was allegedly used by Julius Caesar in his campaign against the Gauls. He took his plaintext and shifted each letter three positions up in the alphabet, wrapping around once he reached the end (so that ‘Z’ would become ‘C’). His correspondents would reverse the process: by moving each letter in the encrypted text down three letters the original message would be recovered. Knowing how to encrypt the text also gave the knowledge of how to decrypt the text: the process wasn’t identical (one shifted up, the other shifted down), but knowing one process the other one could trivially be discovered.

This trait, that of encryption and decryption being two sides of the same coin, is the defining trait of symmetric cryptography. Modern-day symmetric ciphers are much more complex than Caesar’s scheme, but they still work in fundamentally the same way. Knowledge of how to encrypt reveals knowledge of how to decrypt, and vice-versa. The symmetry between those two operations leads to the name “symmetric cryptography”.

Symmetric cryptography is fast, well-studied, and safe. It has one critical drawback, though: you have to have a secure communications channel by which you can share the key with someone. If you already have a secure communications channel, though, do you really need cryptography?

The word ‘key’ is unfortunately ambiguous. It can either refer to the mathematical structures that allow encryption, decryption, signing and verification to occur, or to the rather large blobs of data that contain those mathematical structures as well as information about the person associated with it, additional subkeys, and so forth.

With respect to the large blobs of data, it is preferable to call them ‘certificates’, so that the word ‘key’ may be unambiguously recognized as meaning just the mathematical structures. Unfortunately, this is a custom that seems to be honored mostly in the breach.

A certificate is a large data structure that contains one or more keys, and optionally information that identifies the user, designated revokers, who has vouched for this certificate, and so on.

A keyserver is a service that publishes public-key certificates and makes them searchable. You can upload your certificate to a keyserver so that other users can find it. There are distributed networks of keyservers that share keys, so you only need to upload your key once to that network.

One widely-used keyserver network is sks-keyservers.net. SKS stands for “Synchronizing Key Server”. You can use this network by supplying the --keyserver pool.sks-keyservers.net option.

RSA is the world’s premier asymmetric cryptographic algorithm, and is built on the difficulty of factoring extremely large composites. GnuPG supports RSA with key sizes of between 1024 and 4096 bits.

The United States’ National Institute for Standards and Technology (NIST) established the Digital Signature Algorithm (DSA) as a government standard for digital signatures. Originally, it supported key lengths between 512 and 1024 bits. Recently, NIST has declared 512-bit keys obsolete: now, DSA is available in 1024, 2048 and 3072-bit lengths.

DSA belongs to the Elgamal family of algorithms, and is very well-regarded.

Elgamal may refer to either a family of cryptographic algorithms built around the difficulty of computing discrete logarithms in a finite field, or one particular asymmetric encryption algorithm based on that problem. The former is normally referred to as “the Elgamal family,” and the latter is normally referred to as simply “Elgamal.”

GnuPG supports the Elgamal asymmetric encryption algorithm in key lengths ranging from 1024 to 4096 bits.

There is also an Elgamal signature algorithm, which GnuPG no longer supports.

Leading up to the year 2000, it was obvious that the old Data Encryption Standard (DES) was on its last legs and needed to be replaced. 3DES was available as a stopgap measure, but there was a lot of pressure to make a new encryption standard that made use of the last few decades of cryptologic research.

The United States National Institute of Standards and Technology (NIST) held an open competition to select the new encryption standard. In the summer of 2000, a cipher named Rijndael (pronounced “RAIN-doll”) was selected as the new Advanced Encryption Standard, or AES.

AES is a thoroughly modern cipher design and may be used with confidence.

Blowfish and Twofish are well-regarded symmetric ciphers. Blowfish should not be used to encrypt files larger than 4Gb in size, but Twofish has no such restrictions. These algorithms are modern, and may be used with confidence.

In the 1970s, IBM developed a new symmetric cipher called the Data Encryption Standard (DES). They overdesigned it horribly: even after three decades, the only way to break DES is by brute force. Unfortunately, standard DES has a small enough keyspace to be susceptible to brute-forcing.

A new variant of DES was needed. 3DES, which is made of three DES algorithms running together with three independent keys, was the result. 3DES is ungainly, ugly, slow, and has all the aesthetics of a Soviet workers’ housing bloc. It has also withstood three decades of cryptanalysis and is still going strong.

Due to its 1970s-era 64-bit block size, it should not be used to encrypt more than about 4Gb of data. Beyond that, though, it is solid as a rock, and very few GnuPG users will ever notice a problem with it. Provided you’re not encrypting more than 4Gb of data you may use 3DES with confidence.

Carlisle Adams and Stafford Tavares (the “CA” and the “ST” in “CAST”) developed the CAST algorithm in 1996. It was later approved for Canadian government use.

CAST has many names: CAST, CAST5, CAST5-128 and CAST-128 all refer to the same algorithm.

Internally, CAST is distinctly similar to Blowfish, another well-respected algorithm. Like 3DES, its 64-bit block size means it should not be used to encrypt files larger than 4Gb in size. With that said, though, CAST is a modern cipher and may be used with confidence.

During roughly the same time period that NIST was running the Advanced Encryption Standard trials, Japan’s CRYPTREC and the European Union’s NESSIE were running their own similar trials. Camellia is the cipher that won the NESSIE and CRYPTREC trials, much in the same way that Rijndael won the United States’ AES trials.

Camellia is a thoroughly modern cipher design and may be used with confidence.

The Secure Hash Algorithms are cryptographic hash functions originally devised by the United States’ National Security Agency. The algorithms have been made publicly available and have been subjected to an astonishing amount of peer review.

• SHA and/or SHA-0: the original Secure Hash Algorithm, generating 160-bit outputs. Flaws were discovered in it almost immediately. SHA-0 never gained much traction in the cryptologic community, and it is not present in GnuPG.
• SHA-1: This is SHA-0 with the flaws fixed, and not much else in the way of changes. It still generates 160-bit outputs. SHA-1 has not aged well. Although it is still believed to be safe, it would be advisable to use another, different hash function if possible.
• SHA-224, 256, 384, or 512: This is a massively-overhauled SHA-1 which generates larger hashes (224, 256, 384, or 512 bits). Right now, these are the strongest hashes in GnuPG.
• SHA-3: SHA-3 is a completely new hash algorithm that makes a clean break with the previous SHAs. It is believed to be safe, with no warnings about its usage. It hasn’t yet been officially introduced into the OpenPGP standard, and for that reason GnuPG doesn’t support it. However, SHA-3 will probably be incorporated into the spec, and GnuPG will support it as soon as it does.

MD5 is a 128-bit cryptographic hash function invented by Ron Rivest (the ‘R’ of ‘RSA’) in the early 1990s. For many years it was one of the standard algorithms of the field, but is now completely obsolete. For that reason, MD5 is not supported by GnuPG.

ZLIB, ZIP and BZIP refer to different kinds of compression algorithms. GnuPG will use one of these three algorithms to compress your data before encrypting it, unless GnuPG can see the data is already compressed.

A revocation certificate is a certificate that possesses the information necessary to mark another certificate as unusable. This is called ‘revoking’ the certificate.

We recommended you create a revocation certificate immediately after generating a new GnuPG certificate. Store it somewhere safe. Consult the FAQ instructions on how to do this.

A designated revoker is a person, identified by a certificate, that has the authority to revoke another certificate held by a different person. For instance, if you were using GnuPG in a corporate environment the IT staff might be listed as a designated revoker for your certificate, so that when you left the company the IT staff could revoke your certificate.

Although a certificate makes certain assertions about identity, these assertions cannot be blindly trusted. (Consider, for instance, whether you should trust a certificate that claims to belong to obama@whitehouse.gov.)

If you trust the certificate’s assertions, you are said to have ‘validated’ the certificate. Validation can be done by fiat or as the result of a process. For instance, you validate your own certificate by fiat: “this certificate says it belongs to me, and I trust it.” Validating other certificates, though, should probably have a little more rigor involved. How much rigor will depend entirely on your own particular needs and the threats you face.

‘Trust’ refers to how thoroughly a certificate has been validated. The terms are used somewhat interchangeably.

If a certificate has been validated, and if you trust the person owning that certificate to do proper validation of certificates, you can tell GnuPG “I am willing to trust this person’s validations as if they were my own.”

For instance: Alice has fully validated Bob’s certificate. She further believes, based on her knowledge of Bob, that he will be as careful as she is about the certificates he validates. Alice declares she has ownertrust in Bob. Now, any certificates that Bob validates will appear to Alice as valid, too.

The very first thing is to join the GnuPG-Users mailing list. You’ll find it to be a welcoming community that’s friendly to newcomers and is eager to help out.

No. GnuPG has sensible defaults right out of the box. You don’t need to tune GnuPG before you can use it.

The overwhelming majority of users will be well-served by generating 2048-bit RSA keys. This is the default behavior for GnuPG.

The overwhelming majority of users will be well-served by generating 2048-bit RSA keys. This is the default behavior for GnuPG.

The short answer is, “your computer is doing a lot of work.” But don’t worry: although generating new certificates can take a while, actually using them once they’re made is quite fast.

Generate a revocation certificate, and store it in a safe place. Alternately, you may wish to appoint a designated revoker.

A designated revoker is someone whom you trust to revoke your certificates on your behalf. This person may revoke your certificates without needing a revocation certificate. For instance, you may wish to appoint your lawyer as your designated revoker so that, in the event of your untimely death, your lawyer may revoke your certificates.

To add a revoker, use the following command line:

gpg --edit-key [your key ID here] addrevoker

When prompted, enter the key ID of the person whom you wish to appoint as a revoker. The revoker’s key must be fully validated.

A revocation certificate marks another certificate as unusable.

To generate a revocation certificate for your key, do:

gpg --armor --output revoke.asc --gen-revoke [your key ID]

Copy revoke.asc to a safe place.

gpg --keyserver pool.sks-keyservers.net --send-key [your certificate ID]

You should only upload your own certificates to the keyservers, or obtain the certificate holder’s permission before doing so. In some circles it’s considered rude to upload someone else’s certificate; not everyone wants to publish their key publicly.

GnuPG looks at a file called gpg.conf to determine various runtime parameters. On UNIX systems this file can be found in ~/.gnupg. On Windows systems open Explorer and go to %APPDATA%\Roaming\GnuPG.

The good news is, you really shouldn’t need to. That said, the following is Rob Hansen’s gpg.conf file.

# Tell GnuPG that I want maximum OpenPGP conformance.
openpgp

# Disable a few messages from GnuPG that I know I don’t need.
no-greeting
no-secmem-warning

# Don’t include a version number or a comment in my output.
no-emit-version
no-comments

# Use full 16-character key IDs, not short 8-character key IDs.
keyid-format long

# Use the global keyserver network for certificate lookups.
# Further, whenever I send or receive something to/from the
# keyserver network, clean up what I get or send.
keyserver pool.sks-keyservers.net
keyserver-options import-clean-sigs import-clean-uids export-clean-sigs export-clean-uids

# If I don’t explicitly state which certificate to use, use this one.
default-key 1DCBDC01B44427C7

# Always include signatures from these two certificates.
local-user 1DCBDC01B44427C7

# Always add these two certificates to my recipients list.
encrypt-to 1DCBDC01B44427C7

# Turn "From" into "> From", in order to play nice with UNIX mailboxes.
escape-from-lines

# Prefer strong hashes whenever possible.
personal-digest-preferences SHA256 SHA384 SHA512 SHA224 RIPEMD160

# Prefer more modern ciphers over older ones.
personal-cipher-preferences CAMELLIA256 AES256 TWOFISH CAMELLIA192 AES192 CAMELLIA128 AES BLOWFISH CAST5 3DES

# Turn up the compression level and prefer BZIP2 over ZIP and ZLIB.
bzip2-compress-level 9
compress-level 9
personal-compress-preferences BZIP2 ZIP ZLIB

Many people have had excellent luck with pool.sks-keyservers.net. On OS X, some people have needed to use ipv4.pool.sks-keyservers.net instead.

Commands tell GnuPG what to do: options tell GnuPG how to do it. For instance, encrypt is a command, and armor is an option that tells GnuPG to ensure the output contains only printable characters.

Some of the most commonly used options are:

Produce more output explaining what GnuPG is doing:

-v, --verbose

Make no changes; this is useful for testing a command line that will modify keys or generate output:

-n, --dry-run

Send output to the named file:

-o FILE, --output FILE

Create ASCII-armored output that can be safely e-mailed, instead of binary output:

-a, --armor

When encrypting a message, you will usually supply at least one recipient ID with the recipient option. This option can be supplied multiple times to encrypt a message to multiple recipients:

-r KEYID, --recipient KEYID= specify a recipient ID

GnuPG’s primary functions are to encrypt and decrypt messages, and to sign and verify them. It’s possible to sign without encrypting or encrypt without signing.

Signing a file’s content is done with the -s or --sign commands. A variation is -b or --detach-sign, which produces a separate signature without including the file’s content; this is useful for signing a software archive or other large file. The key to use for the signature can be specified with the local-user setting in your gpg.conf file, or with the -u, --local-user options.

Encrypting a file’s content is done with the -e or --encrypt commands. Recipients are specified with the -r or --recipient options.

GnuPG’s default action is to decrypt and verify its input file, writing the contents to standard output or to the filename specified by the -o or --output options. The --verify command will only verify the signature without writing the file’s contents anywhere.

These commands are the most commonly used. GnuPG has many more commands, largely for managing your keyring containing your private keys and the certificates of others.

In order to send an encrypted message or verify a signature, you must obtain the certificate for the recipient’s/signer’s public key.

Occasionally you might obtain the certificate physically, by meeting the certificate holder face-to-face and exchanging the certificate on some storage medium such as a USB stick, memory card, or portable disk. Or you might download a copy of the certificate from the holder’s web site.

Once obtained in one of these ways, you can add the certificate to your collection of public keys by doing:

gpg --import certificate.txt

More commonly, you’ll download a correspondent’s certificate from a keyserver.

There is also a network of public keyservers, accessible under the collective hostname pool.sks-keyservers.net. GnuPG users can upload their certificates to the keyservers, and other users can then search for and download them.

gpg --keyserver pool.sks-keyservers.net --search [email address, name, key ID, etc.]

GnuPG will list matching certificates and prompt you to select which ones you wish to download and add to your keyring.

People will obtain new signatures for their certificates from time to time. gpg --refresh-keys will recheck all of the certificates on your public key and download any new signatures for those keys.

gpg --keyserver pool.sks-keyservers.net --recv-key [fingerprint]

If you were to receive a letter in the mail that claimed to be from the President of the United States, would you believe it? Probably not, because anyone can put together official-looking letterhead: you’d insist on doing some kind of checking to make sure that no one was fooling with you.

The same applies to email. A certificate can claim to be from anyone. You have to make sure that the certificate really belongs to whom it claims it belongs to. That process of making sure is called ‘validation’.

This advice is controversial.

It’s controversial for a simple reason: every Tom, Dick and Harry has their own idea about the “right way” to validate certificates. Some of these people are well-informed and some of them are just plain unhinged. In the end, you are responsible for making your own decisions. That said, the following is generally agreed upon as being a reasonable procedure:

1. Meet the certificate holder face-to-face.
2. Ask to see two forms of government-issued identification.
3. Upon verifying the person really is who they claim to be, ask this person to provide their certificate’s fingerprint, their email address, and where you can obtain a copy of their certificate. (Example: “My fingerprint is 4541 BB01 8EA4 8F99 19CA 3701 2380 6BE5 D6B9 8E10, and you can find it on pool.sks-keyservers.net.”)
4. On your own computer, retrieve the person’s certificate from the specified location. Check to make sure the email address they gave you is one that’s also listed on the certificate. Check to make sure the fingerprint of the certificate you’ve downloaded matches the fingerprint the person gave you.
5. gpg --edit-key [their certificate ID] sign
6. Once signed, gpg --armor --output signed_cert.asc --export [their certificate ID]
7. Send the file signed_cert.asc to the address they gave you

By following this process you first ensure that you’re speaking to the right person. By comparing the fingerprints of the certificate you have against the fingerprint they specified, you’re ensuring that you have the right certificate. Checking to make sure the email address they gave you is also listed on the certificate is one more check to make sure. Once that’s done, presto, Bob’s your uncle: there’s nothing left to do except sign it and return the newly-signed certificate to the other person.

You encrypted a message to Alice, which means that it requires Alice’s private key to read it. Only Alice has her private key. That’s why you can’t read encrypted traffic you generated: only Alice can read it.

To get around this, add yourself as a recipient (--recipient [your certificate ID]).

Use multiple --recipient options. Remember, options come before commands!

Use multiple --local-user options. Remember, options come before commands!

gpg --armor --recipient [first recipient’s key ID] --local-user [your key ID] --sign --encrypt [filename]

Normally, computers use eight-bit binary code. This often presents trouble for email, which often requires that only printable (seven-bit) characters may be used. By using the --armor flag, GnuPG will generate output containing only printable characters.

An inline signature wraps a textual header and footer around the text to be signed, leaving the text readable without running GnuPG. This doesn’t conceal the text at all and therefore provides no secrecy, but if someone edits the text GnuPG will report that the signature is bad.

To generate an inline signature, run

gpg --armor --output signed_file.asc --local-user [your key ID] --clearsign message_file.txt

To verify the resulting file, simply invoke GnuPG with the filename of the signed file:

gpg signed_file.asc

1. Get a copy of the author’s public certificate and import it to your keyring. It’s important to get the author’s certificate through a trusted source. On the internet, anyone can be pretend to be anyone. Particularly, be careful if the certificate you have doesn’t match the one used for prior code releases.

2. Once you’re confident you have the correct certificate, give it a local signature. Assuming you want to locally sign certificate 1DCBDC01B44427C7, you’d type:

   gpg --edit-key 1DCBDC01B44427C7 lsign

3. Download the software package. Let’s assume it’s called “foo.zip”.

4. Download the detached signature for the package. Let’s assume it’s called “foo.zip.asc”.

5. Run:

   gpg foo.zip.asc

   GnuPG will assume the original file is in foo.zip. (If GnuPG can’t find foo.zip, GnuPG will prompt you for the name of the original package.) If all goes well, GnuPG will report good signatures and you may be confident you’ve received the package as the author intended.

Please note that a good signature doesn’t mean a piece of software is trustworthy, reliable, or bug-free. It just means nobody tampered with it and you’re receiving it as the author intends. Keep a healthy dose of skepticism, and remember that cryptography cannot save us from our own foolishness.

You should use the --batch option. Don’t bother to use a passphrase because there’s usually no way to store it more securely than on the secret keyring itself.

The suggested way to create keys for an automated environment is as follows. First, on a secure machine:

1. If you want to do automatic signing, create a signing subkey for your key. Use the interactive key editing menu by issuing the command:

   gpg --edit-key keyID

   Enter “addkey” and choose whichever key type best suits your needs. (If you don’t know which one is best, choose RSA.)

2. Make sure that you use a passphrase; this is required by the current implementation to let you export the secret key.

3. Run:

   gpg --export-secret-subkeys --no-comment newsubkeyID > secring.auto

4. Copy secring.auto and the public keyring to a test directory.

5. Change to the test directory.

6. Run the command:

gpg --homedir . --edit newsubkeyID

Use the sub-command passwd to remove the passphrase from the subkeys. You may also want to remove all unused subkeys by doing key N and then delkey for each subkey.

1. Copy secring.auto to the target box somehow.

   On the target machine, install secring.auto as the secret keyring and begin writing scripts that invoke GnuPG.

   It’s a good idea to install an intrusion detection system so that you will get notice of a successful intrusion. If that happens, you can revoke all the subkeys installed on that machine and install new subkeys once the machine is secured again.

Check out GPGME (GnuPG Made Easy).

No, nor will there be.

If you received an email claiming to be from a Nigerian oil tycoon, would you believe it? Or would you insist on doing some kind of verification first, in order to make sure that you’re not being scammed or swindled?

The same principle applies here. If you’re using a certificate that claims to belong to Alice, but there’s no evidence it actually belongs to Alice, GnuPG will warn you that you’re using an untrusted certificate.

You probably want to validate the certificate; see this FAQ’s instructions.

GnuPG tries to lock memory so that no other process can see it and so that the memory will not be written to swap. If for some reason it’s not able to do this (for instance, certain platforms don’t support this kind of memory locking), GnuPG will warn you that it’s using insecure memory.

While it’s almost always better to use secure memory, it’s not necessarily a bad thing to use insecure memory. If you own the machine and you’re confident it’s not harboring malware, then this warning can probably be ignored.

GnuPG uses special lines to denote the beginning of a message, the beginning of a signature, and so forth. These lines start with “=–— BEGIN=…”. If your text contains a line beginning with a dash, that line will be slightly mangled in order to prevent GnuPG from misinterpreting your data as one of its special lines.

It’s very hard to give advice on this subject, because everyone will have their own opinion. That said, here are some good guidelines:

• Join the community. Join GnuPG-Users and get involved in the discussions. The conversation is wide-ranging and you’ll encounter a great variety of thoughts and opinions. Reading GnuPG-Users is one of the best ways to educate yourself.
• Practice. If you don’t practice these skills before they become necessary, you won’t be able to use these skills effectively.
• Generate a revocation certificate and keep it safe.
• Use a strong passphrase.
• Keep your computer free of malware.
• Validate certificates correctly.

If someone manages to obtain your secret key, the only thing protecting the key will be your passphrase. A passphrase should be 1) difficult to guess for someone who knows you, and 2) difficult to brute-force by trying every possible combination of characters.

To meet requirement 1), the passphrase shouldn’t be based on publicly-available information about you: your birthday, your spouse’s name, your school’s motto, a line of text from a book, etc. To meet requirement 2), the passphrase should be long: commercially available hardware can try 2.8 billion passwords in a day, which is sufficient to crack a 10-letter all-lowercase password.

One simple approach that produces easy-to-remember passphrases is to generate four to six random words, as illustrated by the XKCD cartoon “Correct, horse! Battery staple!”.

Good places include safe deposit boxes, kept on file with your lawyer, placed in a fireproof safe, and so forth. It should be treated as an important document that needs to be kept safe.

Although there is no guaranteed way of keeping your system free of malware, you can reduce your risk quite a lot by following some basic rules.

1. Keep your system up-to-date. Always apply the latest patches.
2. Stop using old versions of Internet Explorer. If possible, use Mozilla Firefox or Chromium.
3. Don’t open email attachments unless they are expected and come from someone you know.
4. Don’t click on email links unless they are expected and come from someone you know.
5. Be suspicious of requests for personal information, especially if it’s more detail than is strictly necessary to solve a problem.

You can if you want, but it won’t make your private key any more secure. Your private key is already encrypted: your passphrase is the key used to decrypt your private key.

These topics are ‘advanced’ in the sense that you really don’t need to understand them in order to safely and correctly use GnuPG. That said, if you have a more technical question about GnuPG, you may find some of the answers in this section.

Although all the ciphers in GnuPG are believed strong, they are not all equally recommended. For asymmetric ciphers we recommend RSA over DSA and/or Elgamal; for symmetric ciphers we recommend AES, Camellia, and/or Twofish over all the others.

With respect to our RSA recommendation, there is no reason to believe RSA is any better or worse than DSA and/or Elgamal in a cryptographic sense. However, if you ever want to migrate your certificate to a smart card or other cryptographic token, you’ll find RSA is much better supported.

With respect to our symmetric cipher recommendations, we have to explain a little bit about cryptanalysis.

First, ciphers are deterministic: given the same inputs, they generate the same outputs.

Second, ciphers don’t operate on individual bytes. They work on blocks of data, either eight or sixteen bytes large, depending on the cipher.

Third, the OpenPGP standard requires that ciphers run in what’s called a “feedback mode.” In feedback mode, a cipher has two inputs: the random session key used for the message, and the output of the previous block.

Put it all together and imagine what would happen if, within the same message, two identical ciphertext blocks were created. Since the cipher is deterministic (always generates the same output for the same inputs), and since the key and the previous block are the same, the output of this block would be the same. This repetition creates a distinctive pattern which a cryptanalyst might be able to potentially exploit.

For a cipher with an eight-byte block size, you’ll probably repeat a block after about 32 gigabytes of data. This means if you encrypt a single message larger than 32 gigabytes, it’s pretty much a statistical guarantee you’ll have a repeated block. That’s bad. For this reason, we recommend you not use ciphers with eight-byte data blocks if you’re going to be doing bulk encryption. It’s very unlikely you’ll have any problems if you keep your messages under 4 gigabytes in size.

For a cipher with a sixteen-byte block size, you’d need to encrypt a single message that contained more data than is found in the entire internet. In other words, it’s no longer an issue.

Twofish, AES, and Camellia all operate on sixteen bytes at a time. The others all operate on eight bytes at a time.

At the time the decision was made, 2048-bit RSA was thought to provide reasonable security for the next decade or more while still being compatible with the overwhelming majority of the OpenPGP ecosystem.

Largely, yes. According to NIST Special Publication 800-57, published in July 2012, 2048-bit RSA is believed safe until 2030. At present, no reputable cryptographer or research group has cast doubt on the safety of RSA-2048. That said, many are suggesting shifting to larger keys, and GnuPG will be making such a shift in the near future.

In 2014, the German Bundesnetzagentur fuer Elektrizitaet, Gas, Telekommunikation, Post und Eisenbahnen recommended using RSA-2048 for long-term security in electronic signatures.

In 2012, ECRYPT-II published their “Yearly Report on Algorithms and Keysizes” wherein they expressed their belief RSA-1776 will suffice until at least 2020, and RSA-2432 until 2030.

In 2010, France’s Agence Nationale de la Securite des Systems d’Information stated they had confidence in RSA-2048 until at least 2020.

No, although some respected people and groups within the cryptographic community have made such recommendations. Some even recommend 4096-bit keys.

Probably not. The future is elliptical-curve cryptography, which will bring a level of safety comparable to RSA-16384. Every minute we spend arguing about whether we should change the defaults to RSA-3072 or more is one minute the shift to ECC is delayed. Frankly, we think ECC is a really good idea and we’d like to see it deployed as soon as humanly possible.

By all means, feel free to generate certificates with larger keys. GnuPG supports up to 4096-bit keys.

2048-bit RSA is commonly used to secure SSL root signing certificates. It’s also used to sign operating system patches, Authenticode signatures, Java applets and more. RSA-2048 is believed to be safe against attack until at least the year 2030, so use it with confidence.

Because it gives us almost nothing, while costing us quite a lot.

Breaking an RSA-10 key requires you to try each prime number between two and one hundred. There are twenty-five of these, meaning RSA-10 is equivalent to about a 5-bit symmetric cipher. Breaking an RSA-20 key requires you to try each prime number between two and one thousand: there are 168 of them, meaning RSA-20 is equivalent to about an 8-bit cipher. Doubling the keylength (from RSA-10 to RSA-20) didn’t give us the benefit that we naively expected. Each additional bit gives correspondingly less in the way of additional security, and we quickly reach a point of diminishing returns.

That point of diminishing returns happens around RSA-2048. Once you move past RSA-2048, you’re really not gaining very much. At the same time, moving past RSA-2048 means you lose the ability to migrate your certificate to a smartcard, or to effectively use it on some mobile devices, or to interoperate with other OpenPGP applications that don’t handle large keys gracefully.

If you really want a 4096-bit RSA key there’s nothing stopping you: but we sincerely believe the overwhelming majority of users will be well-served with RSA-2048.

Almost always when people use 4096-bit RSA they’re doing so because they believe RSA-4096 to be much stronger than it is. The United States’ National Institute of Standards and Technology (NIST) states that RSA-2048 gives roughly 112 bits of security and RSA-3072 gives roughly 128. There is no formal recommendation on where RSA-4096 lies, but the general consensus is that it would come in somewhere around 140 bits — 28 bits of improvement over RSA-2048. This is an improvement so marginal that it’s really not worth mentioning.

If you need more security than RSA-2048 offers, the way to go would be to switch to elliptical curve cryptography — not to continue using RSA.

RSA-4096 is not a bad idea: it’s just, generally speaking, unnecessary. You gain very little in the way of additional resistance to brute-forcing and cryptanalysis.

No.

The laws of physics require that a certain amount of heat be used in computation. This is a consequence of the Second Law of Thermodynamics, and may not be violated under our current understanding of the laws of physics.

Further, physics requires that a certain amount of time be used in computation. This is a consequence of the Heisenberg Uncertainty Principle, and may not be violated under our current understanding of the laws of physics.

Using these two principles (the Landauer bound and the Margolus–Levitin limit), we can determine quite accurately how much heat would be released by a computer that brute-forced a 128-bit cipher. The results are profoundly silly: it’s enough to boil the oceans and leave the planet as a charred, smoking ruin.

This is not to say that GnuPG cannot be successfully attacked. It is only to say that none of the ciphers in GnuPG are susceptible to brute-forcing.

This depends entirely on what is meant by “successful attack.”

If you mean, “has GnuPG traffic ever been successfully cryptanalyzed?”, the answer is a flat ‘no’. We are unaware of any credible reports of any of the ciphers used in GnuPG having ever been successfully cryptanalyzed.

If you mean, “have people figured out ways to obtain the plaintext anyway?”, the answer is an emphatic ‘yes.’ In a 2007 Drug Enforcement Administration case, a keylogger was installed on a suspect’s computer.

GnuPG protects your traffic against cryptanalysis, but it is not magic fairy dust that can be sprinkled over your data to make it safe against all threats.

Almost certainly. In the past this was a controversial question, but recently there’s come to be a consensus: use PGP/MIME whenever possible. The reason for this is that it’s possible to armor email headers and metadata with PGP/MIME, but sending messages inline leaves this data exposed. As recent years have taught us, the metadata is often as sensitive as the contents of the message. PGP/MIME can protect metadata; inline can’t.

However, please be aware that not all mail servers handle PGP/MIME properly. Some mailing lists are incompatible with it (PGP-Basics, for instance). Some mailing list software mangles PGP/MIME (old versions of Mailman, for instance).

If you have any problems with PGP/MIME, consider carefully whether you need metadata protection. If you don’t, then fall back to inline.

MD5 and SHA-1 should be avoided if possible, and for bulk encryption it’s best to use Camellia, Twofish, or AES. Beyond that guidance there is no “best algorithm” in GnuPG. It’s sort of like asking whether Godzilla or King Kong is better at terrorizing urban cities: there is no clear-cut winner.

This is not to say you shouldn’t have preferences, though. It is only to say that GnuPG’s algorithms are so well-designed for what they do that there is no single “best”. There’s just a lot of personal, subjective choice.

The United States’ National Institute of Standards and Technology (NIST) is responsible for the DSA specification. NIST has not published a 4096-bit DSA variant, and thus GnuPG doesn’t offer it.

The DSA algorithm has gone through several revisions.

GnuPG’s original implementation of DSA supported 1024-bit keys that used either SHA-1 or RIPEMD-160 as hashes.

When the United States’ National Institute of Standards and Technology (NIST) revised the specification to support 2048- and 3072-bit keys, they also required longer hashes be used. DSA-2048 required a 224-bit hash (SHA-224, or a longer hash cut down to 224 bits), and DSA-3072 required a 256-bit hash (SHA-256, or a longer hash cut down to 256 bits). They also now allowed for stronger hashes to be used for DSA-1024: if they were more than 160 bits, they would simply be cut down.

So, depending on how you have GnuPG configured, GnuPG might be forced to use SHA-1 and/or RIPEMD-160 with DSA-1024; GnuPG might be able to use any of the longer SHAs with DSA-1024; GnuPG might use SHA-224, -256, -384 or -512 for DSA-2048; GnuPG might use SHA-256, SHA-384 or SHA-512 for DSA-3072.

PGP 2.6 was released almost twenty-five years ago and is now completely obsolete. We strongly advise against using PGP 2.6 if you have any choice in the matter. Due to PGP 2.6 being obsolete, GnuPG dropped support for it years ago in the GnuPG 2.0 series.

If you absolutely must have PGP 2.6 support, we recommend you use GnuPG's oldest supported version, 1.4, which can still handle PGP 2.6 traffic. We still urge you to migrate your documents to OpenPGP, as we will not be supporting GnuPG 1.4 for much longer.