International Association
for Cryptologic Research

Since Knudsen and Rijmen proposed the known-key attacks in ASIACRYPT 2007, the open-key model becomes more and more popular. As the other component of the open-key model, chosen-key model was applied to the full attacks on AES-256 by Biryukov et al. in CRYPTO 2009. In this paper, we explore how practically the chosen-key model affect the real-world cryptography and show that 11-round generic Feistel-SP block cipher is no longer safe in its hashing modes (MMO and MP mode) as there exist collision attacks. This work improves Sasaki and Yasuda’s collision attacks by 2 rounds with two interesting techniques. First, we for the first time use the available degrees of freedom in the key to reduce the complexity of the inbound phase, which extends the previous 5-round inbound differential to a 7-round one. This results in a 12-round chosen-key distinguisher of Feistel-SP block cipher. Second, inspired by the idea of Wang et al., we construct collisions using two blocks. The rebound attack is used in the second compression function. We carefully balance the freedom of the first block and the complexity of the rebound attack, and extend the chosen-key attack to a 11-round collision attack on its hashing modes (MMO and MP mode).

In this paper, we describe attacks on the recently proposed Haraka hash functions. First, for the two hash functions Haraka-256/256 and Haraka-512/256 in the family, we show how two colliding messages can be constructed in about 216 function evaluations. Second, we invalidate the preimage security claim for Haraka-512/256 with an attack finding one preimage in about 2192 function evaluations. These attacks are possible thanks to symmetries in the internal state that are preserved over several rounds.

A good linear diffusion layer is a prerequisite in the design of block ciphers. Usually it is obtained by combining matrices with optimal diffusion property over the Sbox alphabet. These matrices are constructed either directly using some algebraic properties or by enumerating a search space, testing the optimal diffusion property for every element. For implementation purposes, two types of structures are considered: Structures where all the rows derive from the first row and recursive structures built from powers of companion matrices. In this paper, we propose a direct construction for new recursive-like MDS matrices. We show they are quasi-involutory in the sense that the matrix-vector product with the matrix or with its inverse can be implemented by clocking a same LFSR-like architecture. As a direct construction, performances do not outperform the best constructions found with exhaustive search. However, as a new type of construction, it offers alternatives for MDS matrices design.

The block cipher Kuznyechik and the hash function Streebog were recently standardized by the Russian Federation. These primitives use a common 8-bit S-Box, denoted

Recently, many efficient cryptographic hash function design strategies have been explored, not least because of the SHA-3 competition. These designs are, almost exclusively, geared towards high performance on long inputs. However, various applications exist where the performance on short (fixed length) inputs matters more. Such hash functions are the bottleneck in hash-based signature schemes like SPHINCS or XMSS, which is currently under standardization. Secure functions specifically designed for such applications are scarce. We attend to this gap by proposing two short-input hash functions (or rather simply compression functions). By utilizing AES instructions on modern CPUs, our proposals are the fastest on such platforms, reaching throughputs below one cycle per hashed byte even for short inputs, while still having a very low latency of less than 60 cycles. Under the hood, this results comes with several innovations. First, we study whether the number of rounds for our hash functions can be reduced, if only second-preimage resistance (and not collision resistance) is required. The conclusion is: only a little. Second, since their inception, AES-like designs allow for supportive security arguments by means of counting and bounding the number of active S-boxes. However, this ignores powerful attack vectors using truncated differentials, including the powerful rebound attacks. We develop a general tool-based method to include arguments against attack vectors using truncated differentials.

Statistical attacks form an important class of attacks against block ciphers. By analyzing the distribution of the statistics involved in the attack, cryptanalysts aim at providing a good estimate of the data complexity of the attack. Recently multiple papers have drawn attention to how to improve the accuracy of the estimated success probability of linear key-recovery attacks. In particular, the effect of the key on the distribution of the sample correlation and capacity has been investigated and new statistical models developed. The major problem that remains open is how to obtain accurate estimates of the mean and variance of the correlation and capacity. In this paper, we start by presenting a solution for a linear approximation which has a linear hull comprising a number of strong linear characteristics. Then we generalize this approach to multiple and multidimensional linear cryptanalysis and derive estimates of the variance of the test statistic. Our simplest estimate can be computed given the number of the strong linear approximations involved in the offline analysis and the resulting estimate of the capacity. The results tested experimentally on SMALLPRESENT-[4] show the accuracy of the estimated variance is significantly improved. As an application we give more realistic estimates of the success probability of the multidimensional linear attack of Cho on 26 rounds of PRESENT.

We present an invariant subspace attack on the block cipher Midori64, proposed at Asiacrypt 2015. Our analysis shows that Midori64 has a class of 232 weak keys. Under any such key, the cipher can be distinguished with only a single chosen query, and the key can be recovered in 216 time with two chosen queries. As both the distinguisher and the key recovery have very low complexities, we confirm our analysis by implementing the attacks. Some tweaks of round constants make Midori64 more resistant to the attacks, but some lead to even larger weak-key classes. To eliminate the dependency on the round constants, we investigate alternative S-boxes for Midori64 that provide certain level of security against the found invariant subspace attacks, regardless of the choice of the round constants. Our search for S-boxes is enhanced with a dedicated tool which evaluates the depth of any given 4-bit S-box that satisfies certain design criteria. The tool may be of independent interest to future S-box designs.

AEZ is a parallelizable, AES-based authenticated encryption algorithm that is well suited for software implementations on processors equipped with the AES-NI instruction set. It aims at offering exceptionally strong security properties such as nonce and decryption-misuse resistance and optimal security given the selected ciphertext expansion. AEZ was submitted to the authenticated ciphers competition CAESAR and was selected in 2015 for the second round of the competition. In this paper, we analyse the resilience of the latest algorithm version, AEZ v4.1 (October 2015), against key-recovery attacks. While AEZ modifications introduced in 2015 were partly motivated by thwarting a key-recovery attack of birthday complexity against AEZ v3 published at Asiacrypt 2015 by Fuhr, Leurent and Suder, we show that AEZ v4.1 remains vulnerable to a key-recovery attack of similar complexity and security impact. Our attack leverages the use, in AEZ, of an underlying tweakable block cipher based on a 4-round version of AES. Although the presented key-recovery attack does not violate the security claims of AEZ since the designers made no claim for beyond-birthday security, it can be interpreted as an indication that AEZ does not fully meet the objective of being an extremely conservative and misuse-resilient algorithm.

MDS matrices are used as building blocks of diffusion layers in block ciphers, and XOR count is a metric that estimates the hardware implementation cost. In this paper we report the minimum value of XOR counts of 4 × 4 MDS matrices over F24 and F28 , respectively. We give theoretical constructions of Toeplitz MDS matrices and show that they achieve the minimum XOR count. We also prove that Toeplitz matrices cannot be both MDS and involutory. Further we give theoretical constructions of 4 × 4 involutory MDS matrices over F24 and F28 that have the best known XOR counts so far: for F24 our construction gives an involutory MDS matrix that actually improves the existing lower bound of XOR count, whereas for F28 , it meets the known lower bound.

Real-world applications of authenticated encryption often require the encryption to be computable online, e.g. to compute the ith block of ciphertext after having processed the first i blocks of plaintext. A significant line of research was dedicated to identifying security notions for online authenticated encryption schemes, that capture various security goals related to real-life scenarios. Fouque, Joux, Martinet and Valette proposed definitions of privacy and integrity against adversaries that can query their oracles in a blockwise-adaptive manner, to model memory-constrained applications. A decade later, Fleischmann, Forler and Lucks proposed the notion of online nonce misuse-resistant authenticated encryption (OAE) to capture the security of online authenticated encryption under nonce-reuse. In this work we investigate the relation between these notions. We first recast the blockwise notions of Fouque et al. to make them compatible with online authenticated encryption schemes that support headers. We then show that OAE and the conjunction of the blockwise notions are “almost” equivalent. We identify the missing property on the side of blockwise notions, and formalize it under the name PR-TAG. With PR-TAG being just an auxiliary definition, the equivalence we finally show suggests that OAE and the blockwise model for online authenticated encryption capture essentially the same notion of security.

We show generic attacks on unbalanced Feistel ciphers based on the meet-in-the-middle technique. We analyze two general classes of unbalanced Feistel structures, namely contracting Feistels and expanding Feistels. In both of the cases, we consider the practical scenario where the round functions are keyless and known to the adversary. In the case of contracting Feistels with 4 branches, we show attacks on 16 rounds when the key length k (in bits) is as large as the block length n (in bits), and up to 24 rounds when k = 2n. In the case of expanding Feistels, we consider two scenarios: one, where different nonlinear functions without particular structures are used in the round function, and a more practical one, where a single nonlinear is used but different linear functions are introduced in the state update. In the former case, we propose generic attacks on 13 rounds when k = n, and up to 21 rounds when k = 2n. In the latter case, 16 rounds can be attacked for k = n, and 24 rounds for k = 2n.

The tweakable Even-Mansour construction generalizes the conventional Even-Mansour scheme through replacing round keys by strings derived from a master key and a tweak. Besides providing plenty of inherent variability, such a design builds a tweakable block cipher from some lower level primitive. In the present paper, we evaluate the multi-key security of TEM-1, one of the most commonly used one-round tweakable Even-Mansour schemes (formally introduced at CRYPTO 2015), which is constructed from a single n-bit permutation P and a function f(k, t) linear in k from some tweak space to {0, 1} n. Based on giant component theorem in random graph theory, we propose a collision-based multi-key attack on TEM-1 in the known-plaintext setting. Furthermore, inspired by the methodology of Fouque et al. presented at ASIACRYPT 2014, we devise a novel way of detecting collisions and eventually obtain a memory-efficient multi-key attack in the adaptive chosen-plaintext setting. As important applications, we utilize our techniques to analyze the authenticated encryption algorithms Minalpher (a second-round candidate of CAESAR) and OPP (proposed at EUROCRYPT 2016) in the multi-key setting. We describe knownplaintext attacks on Minalpher and OPP without nonce misuse, which enable us to recover almost all O(2n/3) independent masks by making O(2n/3) queries per key and costing O(22n/3) memory overall. After defining appropriate iterated functions and accordingly changing the mode of creating chains, we improve the basic blockwiseadaptive chosen-plaintext attack to make it also applicable for the nonce-respecting setting. While our attacks do not contradict the security proofs of Minalpher and OPP in the classical setting, nor pose an immediate threat to their uses, our results demonstrate their security margins in the multi-user setting should be carefully considered. We emphasize this is the very first third-party analysis on Minalpher and OPP.

We devise the first closed formula for the number of rounds of a blockcipher with secret components so that these components can be revealed using multiset, algebraic-degree, or division-integral properties, which in this case are equivalent. Using the new result, we attack 7 (out of 9) rounds of Kuznyechik, the recent Russian blockcipher standard, thus halving its security margin. With the same technique we attack 6 (out of 8) rounds of Khazad, the legacy 64-bit blockcipher. Finally, we show how to cryptanalyze and find a decomposition of generic SPN construction for which the inner-components are secret. All the attacks are the best to date.

Online ciphers, in spite of being insecure against an sprp adversary, can be desirable at places because of their ease of implementation and speed. Here we propose a single-keyed inverse-free construction that achieves online sprp security with an optimal number of blockcipher calls. We also include a partial block construction, without requiring any extra key.

Due to the increased use of devices with restricted resources such as limited area size, power or energy, the community has developed various techniques for designing lightweight ciphers. One approach that is increasingly discussed is to use the cipher key that is stored on the device in non-volatile memory not only for the initialization of the registers but during the encryption/decryption process as well. Recent examples are the ciphers Midori (Asiacrypt’15) and Sprout (FSE’15). This may on the one hand help to save resources, but also may allow for a stronger key involvement and hence higher security. However, only little is publicly known so far if and to what extent this approach is indeed practical. Thus, cryptographers without strong engineering background face the problem that they cannot evaluate whether certain designs are reasonable (from a practical point of view) which hinders the development of new designs.In this work, we investigate this design principle from a practical point of view. After a discussion on reasonable approaches for storing a key in non-volatile memory, motivated by several commercial products we focus on the case that the key is stored in EEPROM. Here, we highlight existing constraints and derive that some designs, based on the impact on their throughput, are better suited for the approach of continuously reading the key from all types of non-volatile memory. Based on these findings, we improve the design of Sprout for proposing a new lightweight stream cipher that (i) has a significantly smaller area size than almost all other stream ciphers and (ii) can be efficiently realized using common non-volatile memory techniques. Hence, we see our work as an important step towards putting such designs on a more solid ground and to initiate further discussions on realistic designs.

MANTIS is a lightweight tweakable block cipher published at CRYPTO 2016. In addition to the full 14-round version, MANTIS7, the designers also propose an aggressive 10-round version, MANTIS5. The security claim for MANTIS5 is resistance against “practical attacks”, defined as related-tweak attacks with data complexity 2d less than 230 chosen plaintexts (or 240 known plaintexts), and computational complexity at most 2126−d. We present a key-recovery attack against MANTIS5 with 228 chosen plaintexts and a computational complexity of about 238 block cipher calls, which violates this claim. Our attack is based on a family of differential characteristics and exploits several properties of the lightweight round function and tweakey schedule. To verify the validity of the attack, we also provide a practical implementation which recovers the full key in about 1 core hour using 230 chosen plaintexts.

Quantum computers, that may become available one day, would impact many scientific fields, most notably cryptography since many asymmetric primitives are insecure against an adversary with quantum capabilities. Cryptographers are already anticipating this threat by proposing and studying a number of potentially quantum-safe alternatives for those primitives. On the other hand, symmetric primitives seem less vulnerable against quantum computing: the main known applicable result is Grover’s algorithm that gives a quadratic speed-up for exhaustive search. In this work, we examine more closely the security of symmetric ciphers against quantum attacks. Since our trust in symmetric ciphers relies mostly on their ability to resist cryptanalysis techniques, we investigate quantum cryptanalysis techniques. More specifically, we consider quantum versions of differential and linear cryptanalysis. We show that it is usually possible to use quantum computations to obtain a quadratic speed-up for these attack techniques, but the situation must be nuanced: we don’t get a quadratic speed-up for all variants of the attacks. This allows us to demonstrate the following non-intuitive result: the best attack in the classical world does not necessarily lead to the best quantum one. We give some examples of application on ciphers LAC and KLEIN. We also discuss the important difference between an adversary that can only perform quantum computations, and an adversary that can also make quantum queries to a keyed primitive.

Rotational cryptanalysis is a statistical method for attacking ARX constructions. It was previously shown that ARX-C, i.e., ARX with the injection of constants can be used to implement any function. In this paper we investigate how rotational cryptanalysis is affected when constants are injected into the state. We introduce the notion of an RX-difference, generalizing the idea of a rotational difference. We show how RX-differences behave around modular addition, and give a formula to calculate their transition probability. We experimentally verify the formula using Speck32/64, and present a 7-round distinguisher based on RX-differences. We then discuss two types of constants: round constants, and constants which are the result of using a fixed key, and provide recommendations to designers for optimal choice of parameters.

BLAKE2 is a hash function introduced at ACNS 2013, which has been adopted in many constructions and applications. It is a successor to the SHA-3 finalist BLAKE, which received a significant amount of security analysis. Nevertheless, BLAKE2 introduces sufficient changes so that not all results from BLAKE carry over, meaning new analysis is necessary. To date, all known cryptanalysis done on BLAKE2 has focused on its underlying building blocks, with little focus placed on understanding BLAKE2’s generic security. We prove that BLAKE2’s compression function is indifferentiable from a random function in a weakly ideal cipher model, which was not the case for BLAKE. This implies that there are no generic attacks against any of the modes that BLAKE2 uses.

ChaCha and Salsa are two software oriented stream ciphers that have attracted serious attention in academic as well as commercial domain. The most important cryptanalysis of reduced versions of these ciphers was presented by Aumasson et al. in FSE 2008. One part of their attack was to apply input difference(s) to investigate biases after a few rounds. So far there have been certain kind of limited exhaustive searches to obtain such biases. For the first time, in this paper, we show how to theoretically choose the combinations of the output bits to obtain significantly improved biases. The main idea here is to consider the multi-bit differentials as extension of suitable single-bit differentials with linear approximations, which is essentially a differential-linear attack. As we consider combinations of many output bits (for example 19 for Salsa and 21 for ChaCha), exhaustive search is not possible here. By this method we obtain very high biases for linear combinations of bits in Salsa after 6 rounds and in ChaCha after 5 rounds. These are clearly two rounds of improvement for both the ciphers over the existing works. Using these biases we obtain several significantly improved cryptanalytic results for reduced round Salsa and ChaCha that could not b obtained earlier. In fact, with our results it is now possible to cryptanalyse 6-round Salsa and 5-round ChaCha in practical time.

At CCS 2015, Gueron and Lindell proposed GCM-SIV, a provably secure authenticated encryption scheme that remains secure even if the nonce is repeated. While this is an advantage over the original GCM, we first point out that GCM-SIV allows a trivial distinguishing attack with about 248 queries, where each query has one plaintext block. This shows the tightness of the security claim and does not contradict the provable security result. However, the original GCM resists the attack, and this poses a question of designing a variant of GCM-SIV that is secure against the attack. We present a minor variant of GCM-SIV, which we call GCM-SIV1, and discuss that GCM-SIV1 resists the attack, and it offers a security trade-off compared to GCM-SIV. As the main contribution of the paper, we explore a scheme with a stronger security bound. We present GCM-SIV2 which is obtained by running two instances of GCM-SIV1 in parallel and mixing them in a simple way. We show that it is secure up to 285.3 query complexity, where the query complexity is measured in terms of the total number of blocks of the queries. Finally, we generalize this to show GCM-SIVr by running r instances of GCM-SIV1 in parallel, where r ≥ 3, and show that the scheme is secure up to 2128r/(r+1) query complexity. The provable security results are obtained under the standard assumption that the blockcipher is a pseudorandom permutation.

We introduce subspace trail cryptanalysis, a generalization of invariant subspace cryptanalysis. With this more generic treatment of subspaces we do no longer rely on specific choices of round constants or subkeys, and the resulting method is as such a potentially more powerful attack vector. Interestingly, subspace trail cryptanalysis in fact includes techniques based on impossible or truncated differentials and integrals as special cases. Choosing AES-128 as the perhaps most studied cipher, we describe distinguishers up to 5-round AES with a single unknown key. We report (and practically verify) competitive key-recovery attacks with very low data-complexity on 2, 3 and 4 rounds of AES. Additionally, we consider AES with a secret S-Box and we present a (generic) technique that allows to directly recover the secret key without finding any information about the secret S-Box. This approach allows to use e.g. truncated differential, impossible differential and integral attacks to find the secret key. Moreover, this technique works also for other AES-like constructions, if some very common conditions on the S-Box and on the MixColumns matrix (or its inverse) hold. As a consequence, such attacks allow to better highlight the security impact of linear mappings inside an AES-like block cipher. Finally, we show that our impossible differential attack on 5 rounds of AES with secret S-Box can be turned into a distinguisher for AES in the same setting as the one recently proposed by Sun, Liu, Guo, Qu and Rijmen at CRYPTO 2016

PMAC is a simple and parallel block-cipher mode of operation, which was introduced by Black and Rogaway at Eurocrypt 2002. If instantiated with a (pseudo)random permutation over n-bit strings, PMAC constitutes a provably secure variable input-length (pseudo)random function. For adversaries making q queries, each of length at most l (in n-bit blocks), and of total length σ ≤ ql, the original paper proves an upper bound on the distinguishing advantage of Ο(σ2/2n), while the currently best bound is Ο (qσ/2n).In this work we show that this bound is tight by giving an attack with advantage Ω (q2l/2n). In the PMAC construction one initially XORs a mask to every message block, where the mask for the ith block is computed as τi := γi·L, where L is a (secret) random value, and γi is the i-th codeword of the Gray code. Our attack applies more generally to any sequence of γi’s which contains a large coset of a subgroup of GF(2n). We then investigate if the security of PMAC can be further improved by using τi’s that are k-wise independent, for k > 1 (the original distribution is only 1-wise independent). We observe that the security of PMAC will not increase in general, even if the masks are chosen from a 2-wise independent distribution, and then prove that the security increases to O(q<2/2n), if the τi are 4-wise independent. Due to simple extension attacks, this is the best bound one can hope for, using any distribution on the masks. Whether 3-wise independence is already sufficient to get this level of security is left as an open problem.