Plaintext recovery and tag guessing attacks on authenticated encryption algorithm COLM

Abstract

There are three main approaches related to cryptanalysis of Authenticated Encryption with Associated Data (AEAD) algorithms: Simulating the encryption oracle (universal forgery attack), simulating the decryption oracle (plaintext recovery attack) and producing the valid tag of a given ciphertext (tag guessing attack). In this work, we analyze the security of COLM in these approaches. COLM is one of the AEAD algorithms chosen in the final portfolio for defense-in-depth use case of the CAESAR competition. The ciphers in this portfolio are supposed to provide robust security with their multiple layered defense mechanisms. The main motivation of this work is to examine if COLM indeed satisfies defense-in-depth security. We make cryptanalysis of COLM, particularly in the chosen ciphertext attack (CCA) scenario, once its secret whitening parameter L=EK(0) is recovered. To the best of our knowledge, we give the first example of querying an EME/EMD (Encrypt-linearMix-Encrypt/Decrypt) AEAD scheme in its decryption direction for arbitrary ciphertexts, not produced previously by the oracle, namely either a forgery or tag guessing attack. We construct SEBC/SDBC (Simulation models of the Encryption/Decryption oracles of the underlying Block Cipher) of COLM, thereby forming the first examples of these models of an authenticated EME scheme simultaneously. The combination of our SEBC/SDBC is a powerful tool to mount a universal forgery attack, a tag guessing attack and a plaintext recovery attack. All of these attacks have polynomial time complexities once L is recovered in the offline phase, indicating that the security of COLM against plaintext recovery and tag guessing attacks is limited by the birthday bound. Apart from exploiting SEBC/SDBC, we mount a pair of plaintext recovery attacks and another universal forgery attack. Finally, we make some suggestions to prevent our attacks.