skeeto/enchive

Enchive : encrypted personal archives

Enchive is a tool to encrypt files to yourself for long-term archival. It's a focused, simple alternative to more complex solutions such as GnuPG or encrypted filesystems. Enchive has no external dependencies and is trivial to build for local use. Portability is emphasized over performance.

Supported platforms: Linux, BSD, macOS, Windows

The name is a portmanteau of "encrypt" and "archive," pronounced en'kīv.

Files are secured with ChaCha8, Curve25519, and HMAC-SHA256.

Manual page: enchive(1)

Installation

Clone this repository, then:

$ make PREFIX=/usr install

This will install both the compiled binary and manual page under PREFIX. For staged installs, DESTDIR is also supported. The binary doesn't have any external dependencies and doesn't actually need to be installed before use.

Usage

There are only three commands to worry about: keygen, archive, and extract. The very first thing to do is generate a master keypair using keygen. You will be prompted for the passphrase to protect the secret key, just like ssh-keygen.

$ enchive keygen

By default, this will create two files in $XDG_CONFIG_HOME/enchive (or $HOME/.config/enchive): enchive.pub (public key) and enchive.sec (secret key). On Windows, these are found under %APPDATA%\enchive instead. Distribute enchive.pub to any machines where you plan to archive files. It's sufficient to encrypt files, but not to decrypt them.

To archive a file for storage:

$ enchive archive sensitive.zip

This will encrypt sensitive.zip as sensitive.zip.enchive (leaving the original in place). You can safely archive this wherever.

To extract the file on a machine with enchive.sec, use extract. It will prompt for the passphrase you entered during key generation.

$ enchive extract sensitive.zip.enchive

The original sensitive.zip will be reproduced.

With no filenames, archive and extract operate on standard input and output.

Key management

One of the core features of Enchive is the ability to derive an asymmetric key pair from a passphrase. This means you can store your archive key in your brain! To access this feature, use the --derive (-d) option with the keygen command.

$ enchive keygen --derive

There's an optional argument to --derive that controls the number of key derivation iterations (e.g. --derive=26). The default is 29. This is a power two exponent, so every increment doubles the cost both in memory and computational demands.

If you want to change your protection passphrase, use the --edit option with keygen. It will load the secret key as if it were going to "extract" an archive, then write it back out with the new options. This mode will also regenerate the public key file whether or not it exists.

Enchive has a built-in protection key agent that keeps the protection key in memory for a configurable period of time (default: 15 minutes) after a protection passphrase has been read. This allows many files to be decrypted inside a brief window with only a single passphrase prompt. Use the --agent (-a) global option to enable it. If it's enabled by default, use --no-agent to turn it off.

$ enchive --agent extract file.enchive

Unlike gpg-agent and ssh-agent, this agent need not be started ahead of time. It is started on demand, shuts down on timeout, and does not coordinate with environment variables. One agent is created per unique secret key file. This feature requires a unix-like system.

Notes

The major version number increments each time any of the file formats change, including the key derivation algorithm.

There's no effort at error recovery. It bails out on early on the first error. It should clean up any incomplete files when it does so.

A purposeful design choice is that encrypted/archived files have no distinguishing marks whatsoever (magic numbers, etc.), making them indistinguishable from random data.

Frequently asked questions

This tool will never achieve critical mass, so what's the point?

Enchive doesn't need to interact with any other systems or people, so there's no need for critical mass, nor that there are any other users.

Why can't you use an existing/established tool instead?

I'm not aware of any tool that does everything Enchive does. GnuPG comes close, but doesn't support deriving a key pair from a passphrase. If you're aware of an equal or better tool, please let me know.

Isn't it dangerous to derive a key pair from a passphrase?

It is when it's done incorrectly. However, Enchive uses a memory-hard key derivation scheme that makes cracking passphrases very expensive — prohibitively so for any decent passphrase. This is because anyone who has access to even a single encrypted file can mount an offline attack.

Deriving asymmetric keys from a passphrase is a standard practice in the Bitcoin world: brainwallets. The caveat is that the passphrase must be sufficiently long, preferably chosen by a computer or with dice.

When generating a master key, Enchive's default configuration is extremely paranoid. It would be far cheaper to break into your home and perform an evil maid attack than it would be to crack even a short passphrase. This is not the weak point.

Shouldn't the initialization vector (IV) be generated randomly?

The purpose of an IV is to allow the same key to be safely used multiple times. This is particularly important when the same key is derived on different occasions by Diffie-Hellman between the same key pair. Enchive generates a random ephemeral key pair each time a file is encrypted, so the IV is unnecessary.

Since ChaCha8 requires an IV regardless, Enchive simply uses the hash of the key. This has the additional effect of allowing the client to verify its symmetric key before beginning decryption. Otherwise a wrong key would only be detected by the MAC after decryption has completed.

I'm getting the error "Value too large for defined data type."

This is a flaw in the 32-bit version of glibc that prevents C programs from even opening files larger than 2GB. Compile with "large file support" enabled:

make CFLAGS='-O3 -D_FILE_OFFSET_BITS=64'

Alternatively, use your shell to open files for Enchive:

$ enchive archive <largefile >largefile.enchive

Note that Enchive will not be able to delete shell-opened files in case of errors (tampering, etc.).

Encryption/decryption algorithm

The process for encrypting a file:

1. Generate an ephemeral 256-bit Curve25519 key pair.
2. Perform a Curve25519 Diffie-Hellman key exchange with the master key to produce a shared secret.
3. SHA-256 hash the shared secret to generate a 64-bit IV.
4. Add the format number to the first byte of the IV.
5. Initialize ChaCha8 with the shared secret as the key.
6. Write the 8-byte IV.
7. Write the 32-byte ephemeral public key.
8. Encrypt the file with ChaCha8 and write the ciphertext.
9. Write HMAC(key, plaintext).

The process for decrypting a file:

1. Read the 8-byte ChaCha8 IV.
2. Read the 32-byte ephemeral public key.
3. Perform a Curve25519 Diffie-Hellman key exchange with the ephemeral public key.
4. Validate the IV against the shared secret hash and format version.
5. Initialize ChaCha8 with the shared secret as the key.
6. Decrypt the ciphertext using ChaCha8.
7. Verify HMAC(key, plaintext).

Key derivation algorithm

Enchive uses an scrypt-like algorithm for key derivation, requiring a large buffer of random access memory. Derivation is controlled by a single difficulty exponent D. Secret key derivation requires 512MB of memory (D=29) by default, and protection key derivation requires 32MB by default (D=25). The salt for the secret key is all zeros.

1. Allocate a (1 << D) + 32 byte buffer, M.
2. Compute HMAC_SHA256(salt, passphrase) and write this 32-byte result to the beginning of M.
3. For each uninitialized 32-byte chunk in M, compute the SHA-256 hash of the previous 32-byte chunk.
4. Initialize a byte pointer P to the last 32-byte chunk of M.
5. Compute the SHA-256 hash, H, of the 32 bytes at P.
6. Overwrite the memory at P with H.
7. Take the first D bits of H and use this value to set a new P pointing into M.
8. Repeat from step 5 1 << (D - 5) times.
9. P points to the result.

Compilation

To build on any unix-like system, run make. The resulting binary has no dependencies or external data, so you can just copy/move this into your PATH.

$ make

The easiest way to build with Visual Studio is to use the amalgamation build. On any unix-like system (requires sed):

$ make amalgamation

This will create enchive-cli.c, a standalone C program that you can copy anywhere and compile. Over on Windows:

C:\> cl.exe -nologo -Ox enchive-cli.c advapi32.lib

The compile-time options below also apply to this amalgamation build.

Compile-time configuration

Various options and defaults can be configured at compile time using C defines (-D...).

ENCHIVE_OPTION_AGENT

Whether to expose the --agent and --no-agent option. This option is 0 by default on Windows since agents are unsupported.

ENCHIVE_AGENT_TIMEOUT

The default agent timeout in seconds. This can be configured at run time with an optional argument to --agent.

ENCHIVE_AGENT_DEFAULT_ENABLED

Whether or not to enable the agent by default. This can be explicitly overridden at run time with --agent and --no-agent.

ENCHIVE_PINENTRY_DEFAULT

The default program to use for pinentry.

ENCHIVE_PINENTRY_DEFAULT_ENABLED

Whether or not to use pinentry by default when reading passphrases.

ENCHIVE_FILE_EXTENSION

The file extension to add when archiving and remove when extracting. The default is .enchive, as it appears in the examples.

ENCHIVE_KEY_DERIVE_ITERATIONS

Power-of-two exponent for protection key derivation. Can be configured at run time with --iterations.

ENCHIVE_SECKEY_DERIVE_ITERATIONS

Power-of-two exponent for secret key derivation. Can be configured at run time with the optional argument to --derive.

ENCHIVE_PASSPHRASE_MAX

Maximum passphrase size in bytes, including null terminator.