CoverCrypt

Implementation of the CoverCrypt algorithm which allows creating ciphertexts for a set of attributes and issuing user keys with access policies over these attributes.

• Getting started
• Building and testing
• Features
  • Key generation
  • Serialization
  • Secret key encapsulation
  • Secret key decapsulation
• Benchmarks
• Documentation

Getting started

See examples/runme.rs for a code sample that introduces the main CoverCrypt functionalities. It can be run using cargo run --example runme.

Building and testing

To build the core only, run:

bash cargo build --release

To build everything:

bash cargo build --release --all-features

The code contains numerous tests that you can run using:

bash cargo test --release --all-features

Benchmarks can be run using (one can pass any feature flag):

bash bash ./benches/generate.sh

Features

In CoverCrypt, messages are encrypted using a symmetric scheme. The right management is performed by a novel asymmetric scheme which is used to encapsulate a symmetric key. This encapsulation is stored in an object called encrypted header, along with the symmetric ciphertext.

This design brings several advantages:

• the central authority has a unique key to protect (the master secret key);
• encapsulation can be performed without the need to store any sensitive information (public cryptography);
• encryption is as fast as symmetric schemes can be.

Key generation

Asymmetric keys must be generated beforehand. This is the role of a central authority, which is in charge of:

• generating and updating the master keys according to the right policy;
• generate and update user secret keys.

The CoverCrypt APIs exposes everything that is needed:

• CoverCrypt::setup : generate master keys
• CoverCrypt::join : create a user secret key for the given rights
• CoverCrypt::update : update the master keys for the given policy
• CoverCrypt::refresh : refresh a user secret key from the master secret key

The key generations may be long if the policy contains many rights or if there are many users. But this is usually run once at setup. Key updates and refresh stay fast if the change in the policy is small.

Serialization

The size of the serialized keys and encapsulation is given by the following formulas:

• master secret key:

c 3 * PRIVATE_KEY_LENGTH + LEB128_sizeof(partitions.len()) \ + sum(LEB128_sizeof(sizeof(partition)) + sizeof(partition) + PRIVATE_KEY_LENGTH + 1 [+ INDCPA_KYBER_PRIVATE_KEY_LENGTH])

• public key:

c 2 * PUBLIC_KEY_LENGTH + LEB128_sizeof(partitions.len()) \ + sum(LEB128_sizeof(sizeof(partition)) + sizeof(partition) + PUBLIC_KEY_LENGTH + 1 [+ INDCPA_KYBER_PUBLIC_KEY_LENGTH])

• user secret key:

c 2 * PRIVATE_KEY_LENGTH + LEB128_sizeof(partitions.len()) \ + partition.len() * (PRIVATE_KEY_LENGTH + 1 [+ INDCPA_KYBER_PRIVATE_KEY_LENGTH])

• encapsulation:

c 2 * PUBLIC_KEY_LENGTH + TAG_LENGTH + LEB128_sizeof(partitions.len()) + partition.len() * [INDCPA_KYBER_CIPHERTEXT_LENGTH | PUBLIC_KEY_LENGTH]

• encrypted header (see below):

c sizeof(encapsulation) + DEM_ENCRYPTION_OVERHEAD + sizeof(plaintext)

NOTE: For our implementation CoverCryptX25519Aes256:

• PUBLIC_KEY_LENGTH is 32 bytes
• PRIVATE_KEY_LENGTH is 32 bytes
• TAG_LENGTH is 32 bytes
• DEM_ENCRYPTION_OVERHEAD is 28 bytes (12 bytes for the MAC tag and 16 bytes for the nonce)
• LEB128_sizeof(n) is equal to 1 byte if n is less than 2^7

Secret key encapsulation

This is the core of the CoverCrypt scheme. It allows creating a symmetric key and its encapsulation for a given set of rights.

To ease the management of the encapsulations, an object EncryptedHeaderis provided in the API. An encrypted header holds an encapsulation and a symmetric ciphertext of an optional additional data. This additional data can be useful to store metadata.

Classic implementation sizes:

| Nb. of partitions | Encapsulation size (in bytes) | User decryption key size (in bytes) | |-------------------|-------------------------------|-------------------------------------| | 1 | 131 | 98 | | 2 | 164 | 131 | | 3 | 197 | 164 | | 4 | 230 | 197 | | 5 | 263 | 230 |

Post-quantum implementation sizes:

| Nb. of partitions | Encapsulation size (in bytes) | User decryption key size (in bytes) | |-------------------|-------------------------------|-------------------------------------| | 1 | 1187 | 1250 | | 2 | 2276 | 2435 | | 3 | 3365 | 3620 | | 4 | 4454 | 4805 | | 5 | 5543 | 5990 |

Note: encapsulations grow bigger with the size of the target set of rights and so does the encapsulation time.

Secret key decapsulation

A user can retrieve the symmetric key needed to decrypt a CoverCrypt ciphertext by decrypting the associated EncryptedHeader. This is only possible if the user secret keys contains the appropriate rights.

Benchmarks

The benchmarks presented in this section are run on a Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz.

• CoverCrypt classic implementation
• CoverCrypt post-quantum implementation

Documentation

A formal description and proof of the CoverCrypt scheme is given in this paper. It also contains an interesting discussion about the implementation.

The developer documentation can be found on doc.rs