Superposition Meet-in-the-Middle Attacks: Updates on Fundamental Security of AES-like Ciphers 📺
The Meet-in-the-Middle approach is one of the most powerful cryptanalysis techniques, demonstrated by its applications in preimage attacks on the full MD4, MD5, Tiger, HAVAL, and Haraka-512 v2 hash functions, and key recovery of the full block cipher KTANTAN. The success relies on the separation of a primitive into two independent chunks, where each active cell of the state is used to represent only one chunk or is otherwise considered unusable once mixed. We observe that some of such cells are linearly mixed and can be as useful as the independent ones. This leads to the introduction of superposition states and a whole suite of accompanied techniques, which we incorporate into the MILP-based search framework proposed by Bao et al. at EUROCRYPT 2021 and Dong et al. at CRYPTO 2021, and find applications on a wide range of AES-like hash functions and block ciphers.
Exploring SAT for Cryptanalysis: (Quantum) Collision Attacks against 6-Round SHA-3 📺
In this work, we focus on collision attacks against instances of \shac hash family in both classical and quantum settings. Since the 5-round collision attacks on \shacc-256 and other variants proposed by Guo \etal at JoC~2020, no other essential progress has been published. With a thorough investigation, we identify that the challenges of extending such collision attacks on \shac to more rounds lie in the inefficiency of differential trail search. To overcome this obstacle, we develop a \sat automatic search toolkit. The tool is used in multiple intermediate steps of the collision attacks and exhibits surprisingly high efficiency in differential trail search and other optimization problems encountered in the process. As a result, we present the first 6-round classical collision attack on \shakea with time complexity \cpshake, which also forms a quantum collision attack with quantum time \cpshakeq, and the first 6-round quantum collision attack on \shacc-224 and \shacc-256 with quantum time \cpshattf and \cpshatfs, where $S$ represents the hardware resources of the quantum computer. The fact that classical collision attacks do not apply to 6-round \shacc-224 and \shacc-256 shows the higher coverage of quantum collision attacks, which is consistent with that on SHA-2 observed by Hosoyamada and Sasaki at CRYPTO~2021.
Enhancing Differential-Neural Cryptanalysis 📺
In CRYPTO 2019, Gohr shows that well-trained neural networks can perform cryptanalytic distinguishing tasks superior to traditional differential distinguishers. Moreover, applying an unorthodox key guessing strategy, an 11-round key-recovery attack on a modern block cipher Speck32/64 improves upon the published state-of-the-art result. This calls into the next questions. To what extent is the advantage of machine learning (ML) over traditional methods, and whether the advantage generally exists in the cryptanalysis of modern ciphers? To answer the first question, we devised ML-based key-recovery attacks on more extended round-reduced Speck32/64. We achieved an improved 12-round and the first practical 13-round attacks. The essential for the new results is enhancing a classical component in the ML-based attacks, that is, the neutral bits. To answer the second question, we produced various neural distinguishers on round-reduced Simon32/64 and provided comparisons with their pure differential-based counterparts.
New Techniques for Searching Differential Trails in Keccak 📺
Keccak-f is the permutation used in the NIST SHA-3 hash function standard. Inspired by the previous exhaustive differential trail search methods by Mella et al. at ToSC 2017, we introduce in this paper new algorithms to cover 3-round trail cores with propagation weight at least 53, up from the previous best weight 45. To achieve the goal, the concept of ideal improvement assumption is proposed to construct theoretical representative of subspaces so as to efficiently cover the search space of 3-round trail cores with at least one out-Kernel α state. Of particular note is that the exhaustiveness in 3-round trail core search of at least one out-Kernel α is only experimentally verified. With the knowledge of all 3-round trail cores of weight up to 53, lower bounds on 4/5/6-round trails are tightened to 56/58/108, from the previous 48/50/92, respectively.