Full C implementation of Shufflecake. Shufflecake is a plausible deniability (hidden storage) layer for Linux. https://www.shufflecake.net/
Find a file
2023-07-26 18:57:52 +02:00
benchmark-suite feature/issue-10-benchmarks (#51) 2023-07-26 16:43:14 +00:00
dm-sflc wip:Prepare release v0.4.0 2023-07-23 19:27:02 +02:00
doc doc:Improve header scheme figure 2023-07-24 13:20:14 +02:00
resources/images/badges wip:Prepare release v0.4.0 2023-07-23 19:27:02 +02:00
shufflecake-userland feature/issue-10-benchmarks (#51) 2023-07-26 16:43:14 +00:00
.gitignore fix:Untrack shufflecake binary 2023-07-13 12:02:31 +02:00
AUTHORS refactor:Unified Makefiles 2023-06-06 11:50:19 +02:00
CHANGELOG.md doc:Update SECURITY.md 2023-07-26 18:57:52 +02:00
COPYING Add basic files for license, copyright, etc 2022-12-02 00:59:24 +01:00
COPYRIGHT refactor:Unified Makefiles 2023-06-06 11:50:19 +02:00
LICENSE Initial commit 2022-12-01 23:49:53 +00:00
Makefile feat:Add benchmark (initial skeleton) 2023-07-20 17:49:57 +02:00
README.md feature/issue-10-benchmarks (#51) 2023-07-26 16:43:14 +00:00
SECURITY.md doc:Update SECURITY.md 2023-07-26 18:57:52 +02:00

Status  Version  License  Docs whitepaper  Website  Issue tracker  Mastodon  Please don't upload to GitHub

Shufflecake - Full C Implementation - v0.4.0

Shufflecake is a plausible deniability (hidden storage) layer for Linux. You can consider Shufflecake a spiritual successor of tools like TrueCrypt and VeraCrypt, but vastly improved, both in terms of security and functionality. Official website: https://www.shufflecake.net.

This repository contains source code and documentation to manage Shufflecake volumes.

WARNING: Shufflecake is still experimental software, please do not rely on its security or stability.

Overview

In the context of Shufflecake, a device, or cake, is the underlying raw block device (e.g., a disk) that is formatted to contain hidden data, while a volume, or layer, is the logical, encrypted and hidden "partition" within a device. The device can be a whole USB stick (or disk), a partition, a file-backed loop device, etc. (you likely find it under /dev).

The two operating principles of Shufflecake are:

  • 1 device = multiple volumes
  • 1 volume = 1 password = 1 "secrecy level"

Volumes are password-protected, and embedded in the underlying device as data slices which are indistinguishable from random noise without the proper password. Headers are also indistinguishable from random when not decrypted.

Up to 15 ordered Shufflecake volumes can be created on a single device, with the implicit assumption that "lower-order" volumes (e.g. layer 0) are less secret than "higher-order" ones (e.g. layer 3). The Shufflecake header is designed in such a way that it chains volumes in a backwards-linked list: volume i "points to" (contains the key of) volume i-1. This way, providing the key of volume M allows this tool to traverse the list and also automatically open volumes 0 through M-1 (besides volume M).

Opened volumes appear as block devices of the form sflc_N_M in /dev/mapper. It is up to the user to format them with an appropriate filesystem and mounting them before use. It is recommended to always unlock the most sensitive volume for daily use ("home alone" scenario), even if that volume is not being used or even mounted. Only open a subset of volumes (with a "decoy password") if under coercion. This is due to the high possibility of corrupting hidden volumes otherwise.

For security and consistency reasons, you cannot init/open/close nested volumes within the same device one at a time; the tool only allows to perform these operations on a chain of volumes in a device.

Installation

This implementation of Shufflecake consists of two components: a module for the Linux kernel (dm-sflc), and a shufflecake userland tool. Both are necessary in order to use Shufflecake. They have been tested on Linux kernel versions 6.1 (LTS), 6.2 and 6.3. The following instructions are given for Debian/Ubuntu and similar derivatives.

First of all, you need the kernel headers, device-mapper userspace library, and libgcrypt to compile the source. Use:

sudo apt install linux-headers-$(uname -r) libdevmapper-dev libgcrypt-dev

Important: you have to make sure to install an up-to-date version of libgcrypt that supports the Argon2id KDF algorithm. You need the 1.10.1 or later version, which might not be available in your local repositories (for example, Ubuntu 22.04 LTS), in which case you should find a workaround (either install manually or override your repositories by pinning an up-to-date package). You can check your current version with:

libgcrypt-config --version

After that, just run make. All the compilation artifacts will go in the respective subdirectories, and a copy of the kernel module dm_sflc.ko and the userland tool shufflecake will appear in the root directory. You can clean (delete) all these artifacts with make clean.

For now there is no make install option, if you want you can install Shufflecake system-wide manually, by moving the executable to the appropriate directory (e.g. /usr/bin) and making sure that the kernel module is loaded automatically at boot.

Tests

Limited tests are provided to check pre-install consistency, for now they only test the cryptography layer. Run make test to compile and run them. Remember to make clean afterwards.

Benchmarks

A complete benchmark suite is provided in the benchmark-suite directory, to test the performance of Shufflecake on a given machine and against other tools such as LUKS and VeraCrypt. Please refer to the documentation in that directory for instructions on how to run the benchmark scripts.

Usage

The shufflecake tool requires that the dm-sflc kernel module is already loaded. Therefore, first of all you must load the kernel module with:

sudo insmod dm-sflc.ko

At this point you can run the shufflecake command as sudo shufflecake <action> <block_device>.

When finished, you can remove the module with

sudo rmmod dm-sflc

You can only do this if no Shufflecake volume is open.

Initializing Device

sudo shufflecake init <block_device>

This command creates N volume headers on the given device, each sealed by the respective provided password, by properly formatting and encrypting the first N volume header slots. The number of desired volumes N and the related N passwords will be interactively asked at prompt. Alternatively, you can pass N by command line with:

sudo shufflecake --num-volumes=N init <block_device>

WARNING: If the device is not empty, you will lose all data stored therein. Also, adding additional volumes after initialisation is not yet supported.

This command does not open the volumes (it does not create the virtual devices under /dev/mapper/), it only overwrites the device with randomness and formats the encrypted headers.

Opening Volumes

sudo shufflecake open <block_device>

A password will be required. If this is the password for volume M, then this command will open the first M volumes on the device. These will appear as virtual devices under /dev/mapper/, named sflc_<devID>_0 through sflc_<devID>_(M-1), where devID is a Shufflecake-unique numeric ID assigned by the kernel module (dm_sflc) to the block device.

Notice that this command does not mount the volumes, you have to do it manually. Also notice that at the beginning these volumes will be unformatted; in order to use them to store files, you must first format them with an appropriate filesystem.

Closing Volumes

sudo shufflecake close <block_device>

Closes all the volumes currently open on a device, removing them from /dev/mapper/.

The command only asks for the block device name, i.e., it will close all volumes provided by that device at once, by first forcing a write on disk of any pending changes. However, in order to avoid any risk of data loss, it is always advisable to first umount any mounted open volume.

Advanced Usage

Certain features of Shufflecake are not yet implemented, but sometimes a workaround is possible. Consider all the instructions in this sections as experimental, use them at your own risk.

Volume Corruption Mitigation

As explained in the FAQ, if not all hidden volumes are opened and the user writes data to decoy volumes, there is a risk of data loss on the unopened hidden volumes (so-called "volume corruption"). This is because Shufflecake will try to allocate new slices for the decoy volumes by picking positions at random across all the (unopened) available space. There is ongoing research to mitigate this risk by using some form of redundancy on the volumes, basically by using error-correcting codes that can self-heal from a certain degree of corruption by sacrificing some disk space. In the meantime, users can resort to a "poor man's implementation" of this idea by using RAID redundancy on a partitioned hidden volume. This is still quite "hacky", and at the very minimum should be automated with some scripting.

More in detail:

  1. Open all Shufflecake volumes on a device, e.g. sflc_0_0 and sflc_0_1 in the case of two volumes (one decoy and one hidden).

  2. Partition the volume you want to protect (say, sflc_0_1) into different partitions using, e.g., fdisk or parted. Let's say you obtain 3 equal-sized partitions named sflc_0_11, sflc_0_12 and sflc_0_13.

  3. Use mdadm to build a redundant RAID array on these volumes. For example, if using RAID-1, you should use:

sudo mdadm --create --verbose /dev/md0 --level=1 --raid-devices=3 /dev/mapper/sflc_0_11 /dev/mapper/sflc_0_12 /dev/mapper/sflc_0_13

You can verify the creation of the RAID array with:

cat /proc/mdstat

WARNING: do not save the current RAID configuration to /etc/mdadm/mdadm.conf, otherwise it will super-obviously break plausible deniability.

  1. Format /dev/md0 with a filesystem of your choice, mount it, and use it.

You can use different redundant RAID systems such as RAID-5 (minimum 3 volumes) or RAID-6 (minimum 4 volumes, recommended 5 or more), each of them with different tradeoffs in terms of space wasted, speed, and corruption resilience. In any case there is no 100% guarantee that unopened hidden volumes will not be corrupted when writing data on decoys, so the recommended course of action for the user is still to open all devices for daily use. This method is intended to offer a certain degree of convenience for the user when recovering data in a hidden volume after an adversarial investigation, i.e., having some chance of recovering "gracefully" instead of having to restore from a backup.

Providing Passwords Non-Interactively

By default shufflecake init will ask the user for the number of volumes to create and the sequence of passwords. The number of volumes can be passed as argument with the -n option, but there is currently no built-in mechanism to provide passwords non-interactively, which might be useful, e.g., for scripting. However, there is a simple workaround. As an example with 3 volumes:

echo -e "password1\npassword2\npassword3" | sudo ./shufflecake -n 3 init <block_device>

Even better, you can put password1 ... password3 as the first three lines in a textfile, say passwords.txt, and pass them with:

sudo ./shufflecake -n 3 init <block_device> < passwords.txt

Both methods works with the init action, and we do not have current plans to change this behavior. Notice, however, that there is no guarantee that this will keep working with open or other actions, because the password fetching interface there could be supposed to be more robust, and its implementation might change in the future.

Changelog

Please see the file CHANGELOG.md for a detailed history of changes.

[0.4.0] - 2023-07-24

  • BREAKING CHANGE: slightly modified header field format, removing redundant data and making it adherent to documentation.
  • Implemented reference slice allocation algorithm with much faster performance.
  • Password is now read in a secure shell during open.
  • Added benchmark suite.
  • Fixed bugs in action close and option --skip-randfill.
  • Added doc (currently includes figure of Shufflecake headers structure).

[0.3.1] - 2023-07-15

  • BREAKING CHANGE: fixed parameters for Argon2id KDF, which were previously set too high, resulting in slow opening for non-recent devices. Unfortunately this breaks header format compatibility with device headers initialized with v0.3.0, so please treat v0.3.0 as bugged.
  • Interactive test for Argon2id KDF.
  • Added in README.md a description for a manual procedure for increasing volume resistance to corruption using RAID.

Bugs

Bugs and other issues are tracked at https://codeberg.org/shufflecake/shufflecake-c/issues, please report new bugs by submitting new issues there. Alternatively, you can send an email to bugs@shufflecake.net.

Outstanding known bugs:

  • There is no garbage collection currently implemented, so slices remain flagged as used even if the data therein is deleted. Work is in progress to fix this.

Maintainers

Elia Anzuoni <elianzuoniATgmail.com>

Tommaso "Tomgag" Gagliardoni <tommasoATgagliardoni.net>

Copyright The Shufflecake Project Authors (2022)

Copyright The Shufflecake Project Contributors (2022)

Copyright Contributors to the The Shufflecake Project.

See the AUTHORS file at the top-level directory of this distribution and at https://www.shufflecake.net/permalinks/shufflecake-c/AUTHORS.

The program shufflecake-c is part of the Shufflecake Project. Shufflecake is a plausible deniability (hidden storage) layer for Linux. See https://www.shufflecake.net.

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see https://www.gnu.org/licenses/.