summaryrefslogtreecommitdiff
path: root/Documentation/SECURITY.md
diff options
context:
space:
mode:
authorJakob Unterwurzacher2015-12-01 23:02:12 +0100
committerJakob Unterwurzacher2015-12-01 23:06:56 +0100
commit026ab56de4b61df634ee22089c6e4ed361e3ae98 (patch)
tree4300df342204ef72aecb93f1f10efd4821a9b84a /Documentation/SECURITY.md
parent6515c8d42cfe6b448a73ef5811b3a14c8363ea6c (diff)
Cut down the text in SECURITY.md, add graphs
Diffstat (limited to 'Documentation/SECURITY.md')
-rw-r--r--Documentation/SECURITY.md102
1 files changed, 26 insertions, 76 deletions
diff --git a/Documentation/SECURITY.md b/Documentation/SECURITY.md
index 4db4c24..e9e1015 100644
--- a/Documentation/SECURITY.md
+++ b/Documentation/SECURITY.md
@@ -1,92 +1,42 @@
GoCryptFS Security
==================
-"Security" can be split into "Confidentiality" and "Integrity". The
-security level gocryptfs provides for each is discussed in the next
-sections.
-
-Confidentiality
----------------
-
-Confidentiality means that information cannot be extracted from the
-encrypted data unless you know the key.
-
-### File Contents
-
-* All file contents (even the last bytes) are encrypted using AES-256-GCM
- * This is unbreakable in the foreseeable future. Attacks will focus on
- cracking the password instead (see section "Master Key Storage").
-* Files are segmented into 4096 byte blocks
-* Each block gets a fresh random 96 bit IV (none) each time it is written.
- * This means that identical blocks can not be identified
-
-### File Names
-
-* File names are encrypted using AES-256-CBC with a per-directory IV
-* Each directory get a random 128 bit IV on creation
- * Files with the same name in different directories are encrypted to
- different filenames and can not be identified
-* File names are padded to multiples of 16 bytes
- * This means that the exact length of the name is hidden, only length
- ranges (1-16 bytes, 17-32 bytes etc.) can be determined from the encrypted
- files
+Master Key Storage
+------------------
-### Metadata
+The master key is used to perform content and file name encryption.
+It is stored in `gocryptfs.conf`, encrypted with AES-256-GCM using the
+Key Encryption Key (KEK).
-* The size of the file is not hidden. The exact file size can be calculated
- from the size of the encrypted file.
-* File owner, file permissions and timestamps are not hidden.
+The KEK is generated from the user password using `scrypt`.
-Integrity
----------
+![](https://rawgit.com/rfjakob/gocryptfs/master/Documentation/master-key.svg)
-Integrity means that the data cannot be modified in a meaningful way
-unless you have the key. The opposite of integrity is *malleability*.
+File Contents
+-------------
-### File Contents
+All file contents are encrypted using AES-256-GCM.
-* The used encryption, AES-256-GCM, is a variant of
- *authenticated encryption*. Each block gets a 128 bit authentication
- tag (GMAC) appended.
- * This means that any modification inside a block will be detected when reading
- the block and decryption will be aborted. The failure is logged and an
- I/O error is returned to the user.
-* Every file has a header that contains a 16-byte random *file id*
-* Each block uses the file id and its block number as GCM *authentication data*
- * This means the position of the blocks is protected as well. The blocks
- can not be reordered or copied between different files without
- causing an decryption error.
-* For technical reasons (sparse files), the special "all-zero" block is
- always seen as a valid block that decrypts to all-zero plaintext.
- * This means that whole blocks can be zeroed out
+Files are segmented into 4KB blocks. Each block gets a fresh random
+96 bit IV each time it is modified. A 128-bit authentication tag (GHASH)
+protects each block from modifications.
-### File Names
+Each file has a header containing a random 128-bit file ID. The
+file ID and the block number are mixed into the GHASH as
+*additional authenticated data*. The prevents blocks from being copied
+between or within files.
-* File names are only weakly protected against modifications.
- * Changing a single byte causes a decode error in most of the
- cases. The failure is logged and the file is no longer visible in the
- directory.
- * If no decode error is triggered, at least 16 bytes of the filename will
- be corrupted (randomized).
-* However, file names can always be truncated to multiples of 16 bytes.
+![](https://rawgit.com/rfjakob/gocryptfs/master/Documentation/file-content-encryption.svg)
-### Metadata
+To support sparse files, all-zero blocks are accepted and passed through
+unchanged.
-* The file size is not protected against modifications
- * However, the block integrity protection limits modifications to block
- size granularity.
- * This means that files can be truncated to multiples of 4096 bytes.
-* Ownership, timestamp and permissions are not protected and can be changed
- as usual.
+File Names
+----------
-Master Key Storage
-------------------
+Every directory gets a 128-bit directory IV that is stored in each
+directory as `gocryptfs.diriv`.
-The *master key* is used to perform content and file name encryption.
-It is stored in `gocryptfs.conf`, encrypted with AES-256-GCM using the
-*unlock key*.
+File names are encrypted using AES-256-CBC with the directory IV.
-The unlock key is generated from a user password using `scrypt`.
-A successful decryption of the master key means that the GMAC authentication
-passed and the password is correct. The master key is then used to
-mount the filesystem.
+![](https://rawgit.com/rfjakob/gocryptfs/master/Documentation/file-name-encryption.svg)