An integral distinguisher for a block cipher is defined by a nontrivial subset of plaintexts for which the bitwise sum of (parts of) a certain internal state is independent of the secret key. Such a distinguishing property can be turned into a key-recovery procedure by partially decrypting the ciphertexts under all possible keys and then filtering the key candidates using the integral distinguisher. The behavior of this filter has never been analyzed in depth, and we show that the ubiquitous hypothesis about its behavior is incorrect. Fortunately, the deviation is either limited or can be lifted to improve the underlying attacks. By algorithmically determining the exact subspaces of key candidates to be guessed – whose dimensions are often lower than expected – we are able to improve upon the best known integral key-recovery attacks on various ciphers.

We present a full key-recovery ciphertext-only attack on the tweakable block cipher Halfloop-24 which is standardized by the US military and NATO for use in the Automatic Link Establishment protocol for high-frequency radio. Our attack is based on the observation that the probability-one differential, which was used for known- and chosen-plaintext attacks in prior works, can be viewed as a probability-p differential when the plaintexts are unknown, where p is the network-dependent probability that plaintext differences line up with tweak differences. Further, due to the fact that the tweak is publicly known, the specific values of the plaintexts are not necessary to identify the correct key. Our attack has a theoretical complexity of roughly 2^65 encryptions, given a favorable ALE network setup that allows us to collect enough suitable ciphertexts. To give a hardware-grounded estimate of the cost of the whole attack, we provide optimized CPU and GPU implementations.

End-to-end deep-learning side-channel attacks on public-key implementations have recently become possible even for million-sample traces. However, existing methods require large computational resources and extract only partial key shares, which means that dedicated post-processing is required to turn detected leakage into demonstrations of successful key recovery attacks. We present an end-to-end sequence-to-sequence prediction approach to recover complete 256-bit key shares from single raw traces on the SCAAML ECC datasets recently studied by Bursztein et al. [3]. Our solution combines aggressive trace compression for dimensionality reduction with a 1-D U-Net trained using Connectionist Temporal Classification loss. The key idea is to decouple detecting leakage from mapping each leakage site to the correct part of the secret: the network outputs an annotated map of the trace marking likely leakage sites, and a greedy decoder reconstructs the ordered key bits from that map. Using synthetic tasks, we show that this division of labor circumvents a fundamental problem that causes neural network architectures and training methods commonly used in side-channel analysis to struggle with massive multi-target or misaligned extraction tasks. As a result, we are able to train a single extractor that achieves high accuracy on all four SCAAML ECC datasets in a single training run that takes minutes on a single GPU. The resulting extractors are robust, essentially maintaining their performance under large misalignment (we empirically tested rotations up to 61% of trace length), while degrading gracefully under a variety of trace corruptions, and even time reversal. They transfer across key shares and datasets with little degradation and no retraining. The U-Net outputs also yield prediction maps that localize leakage along the trace prior to decoding (Supplementary code and data for this work are available in the repository https://github.com/agohr/seq2seq-sca ).

We prove that 4r + 4 rounds of an AES variant with independent and uniform random round keys are ε -close to pairwise independent with ε = 2^14 2^-40r . This result follows from a near-optimal bound for a two-norm version of pairwise independence for the Shark construction, depending on the third singular value of the difference-distribution table of the S-boxes. Our analysis combines insights from cryptanalysis—in particular, truncated differentials—and linear algebra over the reals.

In this work, we present new cryptanalytic attacks on recently proposed, theory-inspired constructions of weak pseudorandom functions (weak-PRFs). We demonstrate attacks on several such designs, showing that the initial security arguments require significant refinement. Methodologically, our approach relies on novel observations about the structure of cyclic matrices, applications of Wagner’s generalized birthday technique, and conversion into polynomial systems over F3. These findings highlight the need for a more careful analysis of those weak-PRF candidates.

Most modern cryptanalysis results are obtained through theoretical analysis, often relying on simplifications and idealized assumptions. In this work, we use the parallel computational power of GPUs to experimentally verify a small portion of the cryptanalysis results that have been published in recent years. Our focus is on the ciphers Ascon, ChaCha, and Serpent. In none of the attacks we considered did the theoretical estimates fully match the actual practical values. More precisely, we show that the 4.5-round truncated differential with probability one, the 6-round differential-linear (DL), and the 6-round impossible differential distinguishers on Ascon, as well as the best known 7-and 7.5-round DL distinguisher on ChaCha, do not actually work in practice. Moreover, we demonstrate that the best known 10, 11, and 12-round DL attacks on Serpent perform better in practice than previously estimated. Additionally, we provide a new experimentally obtained 9-round DL distinguisher on Serpent, which can be used in 10 and 11-round attacks with reduced data complexity. In a broader sense, we recommend that cryptanalysts experimentally verify reduced versions of their theoretically obtained analysis results whenever possible. In order to simplify this process, we make our optimized code for the ciphers treated here available for future use.

Key lengths in symmetric cryptography are determined with respect to the brute force attacks with current technology. While nowadays at least 128-bit keys are recommended, there are many standards and real-world applications that use shorter keys. In order to estimate the actual threat imposed by using those short keys, precise estimates for attacks are crucial. In this work we provide optimized implementations of several widely used algorithms on GPUs, leading to interesting insights on the cost of brute force attacks on several real-word applications. In particular, we optimize KASUMI (used in GPRS/GSM), SPECK (used in RFID communication), and TEA3 (used in TETRA). Our best optimizations allow us to try 235.72, 236.72, and 234.71 keys per second on a single RTX 4090 GPU. Those results improve upon previous results significantly, e.g. our KASUMI implementation is more than 15 times faster than the optimizations given in the CRYPTO'24 paper [ACC+24] improving the main results of that paper by the same factor. With these optimizations, in order to break GPRS/GSM, RFID, and TETRA communications in a year, one needs around 11, 22 billion, and 1.36 million RTX 4090 GPUs, respectively. For KASUMI, the time-memory trade-off attacks of [ACC+24] can be performed with 142 RTX 4090 GPUs instead of 2400 RTX 3090 GPUs or, when the same amount of GPUs are used, their table creation time can be reduced to 20.6 days from 348 days, crucial improvements for real world cryptanalytic tasks.

Integral attacks exploit structural weaknesses in symmetric cryptographic primitives by analyzing how subsets of inputs propagate to produce outputs with specific algebraic properties. For the case of (XOR) key-alternating block ciphers using (independent) round keys, at ASIACRYPT'21, Hebborn et al. established the first non-trivial lower bounds on the number of rounds required for ensuring integral resistance in a quite general sense. For the case of adding keys by modular addition, no security arguments are known so far. Here, we present a unified framework for analyzing the integral resistance of primitives using (word-wise) modular addition for key whitening, allowing us to not only fill the gap for security arguments, but also to overcome the heavy computational cost inherent in the case of XOR-whitening.

We present a number of observations on TEA-3, a stream cipher used in TETRA radio networks that was kept secret until recently. While the same also holds for the six other TETRA encryption algorithms, we pick TEA-3 to start with, as (i) it is not obviously weakened as TEA-{1,4,7} but (ii) in contrast to TEA-2 it is approved for extra-European emergency service, and (iii) as already noted by [MBW23] the TEA-3 design surprisingly contains a non-bijective S-box. Most importantly, we show that the 80-bit non-linear feedback shift register operating on the key decomposes into a cascade of two 40-bit registers. Although this hints at an intentional weakness at first glance, we are not able to lift our results to a practical attack. Other than that, we show how the balanced non-linear feedback functions used in the state register of TEA-3 can be constructed.

We study the problem of embedded code encryption, i.e., encryption for binary software code for a secure microcontroller that is stored in an insecure external memory. As every single instruction must be decrypted before it can be executed, this scenario requires an extremely low latency decryption. We present a formal treatment of embedded code encryption security definitions, propose three constructions, namely ACE1, ACE2 and ACE3, and analyze their security. Further, we present ChiLow, a family of tweakable block ciphers and a related PRF specifically designed for embedded code encryption. At the core of ChiLow, there is ChiChi, a new family of non-linear layers of even dimension based on the well-known χ function. Our fully unrolled hardware implementation of ChiLow, using the Nangate 15nm Open Cell Library, achieves a decryption latency of less than 280 picoseconds.

The only known example of an almost perfect nonlinear (APN) permutation in even dimension was obtained by applying CCZ-equivalence to a specific quadratic APN function. Motivated by this result, there have been numerous recent attempts to construct new quadratic APN functions. Currently, 32,892 quadratic APN functions in dimension 8 are known and two recent conjectures address their possible total number. The first, proposed by Y. Yu and L. Perrin (Cryptogr. Commun. 14(6): 1359-1369, 2022), suggests that there are more than 50,000 such functions. The second, by A. Polujan and A. Pott (Proc. 7th Int. Workshop on Boolean Functions and Their Applications, 2022), argues that their number exceeds that of inequivalent quadratic (8,4)-bent functions, which is 92,515. We computationally construct 3,775,599 inequivalent quadratic APN functions in dimension 8 and estimate the total number to be about 6 million.

Key lengths in symmetric cryptography are determined with respect to the brute force attacks with current technology. While nowadays at least 128-bit keys are recommended, there are many standards and real-world applications that use shorter keys. In order to estimate the actual threat imposed by using those short keys, precise estimates for attacks are crucial. In this work we provide optimized implementations of several widely used algorithms on GPUs, leading to interesting insights on the cost of brute force attacks on several real-word applications. In particular, we optimize KASUMI (used in GPRS/GSM), SPECK (used in RFID communication), and TEA3 (used in TETRA). Our best optimizations allow us to try 235.72, 236.72, and 234.71 keys per second on a single RTX 4090 GPU. Those results improve upon previous results significantly, e.g. our KASUMI implementation is more than 15 times faster than the optimizations given in the CRYPTO’24 paper [ACC+24] improving the main results of that paper by the same factor. With these optimizations, in order to break GPRS/GSM, RFID, and TETRA communications in a year, one needs around 11, 22 billion, and 1.36 million RTX 4090 GPUs, respectively. For KASUMI, the time-memory trade-off attacks of [ACC+24] can be performed with 142 RTX 4090 GPUs instead of 2400 RTX 3090 GPUs or, when the same amount of GPUs are used, their table creation time can be reduced to 20.6 days from 348 days, crucial improvements for real world cryptanalytic tasks.

Recently, Baudrin et al. analyzed a special case of Wagner’s commutative diagram cryptanalysis, referred to as commutative cryptanalysis. For a family (E_k)_k of permutations on a finite vector space G, commutative cryptanalysis exploits the existence of affine permutations A,B :G → G , I ∉{A,B} such that E_k ∘ A (x) = B ∘ E_k(x) holds with high probability, taken over inputs x, for a significantly large set of weak keys k. Several attacks against symmetric cryptographic primitives can be formulated within the framework of commutative cryptanalysis, most importantly differential attacks, as well as rotational and rotational-differential attacks. Besides, the notion of c-differentials on S-boxes can be analyzed as a special case within this framework. We discuss the relations between a general notion of commutative cryptanalysis, with A and B being arbitrary functions over a finite Abelian group, and differential cryptanalysis, both from the view of conducting an attack on a symmetric cryptographic primitive, as well as from the view of a theoretical study of cryptographic S-boxes.

While masking is a widely used defense against passive side-channel attacks, its secure implementation in hardware continues to be a manual, complex, and error-prone process. This paper introduces INDIANA, a comprehensive security verification methodology for hardware masking. Our results include a hardware verification tool, enabling a complete analysis of simulation-based security in the glitch-extended probing model and intra-cycle estimations for leakage probabilities in the random probing model. Notably, INDIANA is the first framework to analyze arbitrary masked circuits in both models, even at the scale of full SPN cipher rounds (e.g., AES), while delivering exact verification results. To achieve accurate and comprehensive verification, we propose a partitionable probing distinguisher that allows for fast validation of probe tuples, surpassing current methods that rely on statistical independence. Furthermore, our approach naturally supports extensions to the random probing model by utilizing Fast Fourier-Hadamard Transformations (FHTs). Benchmark results show that INDIANA competes effectively with leading probing model verification tools, such as ironMask, maskVerif, and VERICA. INDIANA is also the first tool that is capable to provide intra-cycle estimations of random probing leakage probabilities for large-scale masked circuits.

Most modern cryptanalysis results are obtained through theoretical analysis, often relying on simplifications and idealized assumptions. In this work, we use the parallel computational power of GPUs to experimentally verify a small portion of the cryptanalysis results that have been published in recent years. Our focus is on the ciphers Ascon, ChaCha, and Serpent. In none of the attacks we considered did the theoretical estimates fully match the actual practical values. More precisely, we show that the 4.5-round truncated differential with probability one, the 6-round differential-linear (DL), and the 6-round impossible differential distinguishers on Ascon, as well as the best known 7- and 7.5-round DL distinguisher on ChaCha, do not actually work in practice. Moreover, we demonstrate that the best known 10, 11, and 12-round DL attacks on Serpent perform better in practice than previously estimated. Additionally, we provide a new experimentally obtained 9-round DL distinguisher on Serpent, which can be used in 10 and 11-round attacks with reduced data complexity. In a broader sense, we recommend that cryptanalysts experimentally verify reduced versions of their theoretically obtained analysis results whenever possible. In order to simplify this process, we make our optimized code for the ciphers treated here available for future use.

At Eurocrypt'25, Belkheyar et al. introduced a new non-linear layer called ChiChi or chi chi. ChiChi is built based on Daemen's chi function, but crucially gives a permutation in even dimension divisible by four. In this work, we generalize their construction in multiple ways, prove their open conjecture regarding the algebraic degree of the inverse of ChiChi, and investigate the properties of it against key recovery attacks.

We present the first comprehensive approach for detecting and analyzing symmetric cryptographic primitives in gate-level descriptions of hardware. To capture both ASICs and FPGAs, we model the hardware as a directed graph, where gates become nodes and wires become edges. For modern chips, those graphs can easily consist of hundreds of thousands of nodes. More abstractly, we find subgraphs corresponding to cryptographic primitives in a potentially huge graph, the sea-of-gates, describing an entire chip. As we are particularly interested in unknown cryptographic algorithms, we cannot rely on searching for known parts such as S-boxes or round constants. Instead, we are looking for parts of the chip that perform highly local computations. A major result of our work is that many symmetric algorithms can be reliably located and sometimes even identified by our approach, which we call HAWKEYE. Our findings are verified by extensive experimental results, which involve SPN, ARX, Feistel, and LFSR-based ciphers implemented for both FPGAs and ASICs. We demonstrate the real-world applicability of HAWKEYE by evaluating it on OpenTitan's Earl Grey chip, an open-source secure micro-controller design. HAWKEYE locates all major cryptographic primitives present in the netlist comprising 424 341 gates in 44.3 s.