International Association
for Cryptologic Research

We study sleepy consensus in the known participation model, where replicas are aware of the minimum number of awake honest replicas. Compared to prior works that almost all assume the unknown participation model, we provide a fine-grained treatment of sleepy consensus in the known participation model and show some interesting results. First, we present a synchronous atomic broadcast protocol with $5\Delta+2\delta$ expected latency and $2\Delta+2\delta$ best-case latency, where $\Delta$ is the bound on network delay and $\delta$ is the actual network delay. In contrast, the best-known result in the unknown participation model (MMR, CCS 2023) achieves $14\Delta$ latency, more than twice the latency of our protocol. Second, in the partially synchronous network (the value of $\Delta$ is unknown), we show that without changing the conventional $n \geq 3f+1$ assumption, one can only obtain a secure sleepy consensus by making the stable storage assumption (where replicas need to store intermediate consensus parameters in stable storage). Finally, still in the partially synchronous network but not assuming stable storage, we prove the bounds on $n \geq 3f+2s+1$ without the global awake time (GAT) assumption (all honest replicas become awake after GAT) and $n \geq 3f+s+1$ with the GAT assumption, where $s$ is the maximum number of honest replicas that may become asleep simultaneously. Using these bounds, we transform HotStuff (PODC 2019) into a sleepy consensus protocol via a timeoutQC mechanism and a low-cost recovery protocol.

Secure transformer inference has emerged as a prominent research topic following the proliferation of ChatGPT. Existing solutions are typically interactive, involving substantial communication load and numerous interaction rounds between the client and the server.

In this paper, we propose NEXUS the first non-interactive protocol for secure transformer inference, where the client is only required to submit an encrypted input and await the encrypted result from the server. Central to NEXUS are two innovative techniques: SIMD ciphertext compression/decompression, and SIMD slots folding. Consequently, our approach achieves a speedup of 2.8$\times$ and a remarkable bandwidth reduction of 368.6$\times$, compared to the state-of-the-art solution presented in S&P '24.

The implementation security of post-quantum cryptography (PQC) algorithms has emerged as a critical concern with the PQC standardization process reaching its end. In a side-channel-assisted chosen-ciphertext attack, the attacker builds linear inequalities on secret key components and uses the belief propagation (BP) algorithm to solve. The number of inequalities leverages the query complexity of the attack, so the fewer the better. In this paper, we use the PQC standard algorithm Kyber512 as a study case to construct bilateral inequalities on key variables with substantially narrower intervals using a side-channel-assisted oracle. The number of such inequalities required to recover the key with probability 1 utilizing the BP algorithm is reduced relative to previous unilateral inequalities. Furthermore, we introduce strategies aimed at further refining the interval of inequalities. Diving into the BP algorithm, we discover a measure metric named JSD-metric that can gauge the tightness of an inequality. We then develop a heuristic strategy and a machine learning-based strategy to utilize the JSD-metrics to contract boundaries of inequalities even with fewer inequalities given, thus improving the information carried by the system of linear inequalities. This contraction strategy is at the algorithmic level and has the potential to be employed in all attacks endeavoring to establish a system of inequalities concerning key variables.

The conventional Byzantine fault tolerance (BFT) paradigm requires replicated state machines to execute deterministic operations only. In practice, numerous applications and scenarios, especially in the era of blockchains, contain various sources of non-determinism. Despite decades of research on BFT, we still lack an efficient and easy-to-deploy solution for BFT with non-determinism—BFT-ND, especially in the asynchronous setting. We revisit the problem of BFT-ND and provide a formal and asynchronous treatment of BFT-ND. In particular, we design and implement Block-ND that insightfully separates the task of agreeing on the order of transactions from the task of agreement on the state: Block-ND allows reusing existing BFT implementations; on top of BFT, we reduce the agreement on the state to multivalued Byzantine agreement (MBA), a somewhat neglected primitive by practical systems. Block-ND is completely asynchronous as long as the underlying BFT is asynchronous. We provide a new MBA construction significantly faster than existing MBA constructions. We instantiate Block-ND in both the partially synchronous setting (with PBFT, OSDI 1999) and the purely asynchronous setting (with PACE, CCS 2022). Via a 91-instance WAN deployment on Amazon EC2, we show that Block-ND has only marginal performance degradation compared to conventional BFT.

Symmetric cryptography primitives are constructed by iterative applications of linear and nonlinear layers. Constructing efficient circuits for these layers, even for the linear one, is challenging. In 1997, Paar proposed a heuristic to minimize the number of XORs (modulo 2 addition) necessary to implement linear layers. In this study, we slightly modify Paar’s heuristics to find implementations for nonlinear Boolean functions, in particular to homogeneous Boolean functions. Additionally, we show how this heuristic can be used to construct circuits for generic Boolean functions with small number of AND gates, by exploiting affine equivalence relations.

Although having been popular for a long time, Byzantine Fault Tolerance (BFT) consensus under the partially-synchronous network is denounced to be inefficient or even infeasible in recent years, which calls for a more robust asynchronous consensus. On the other hand, almost all the existing asynchronous consensus are too complicated to understand and even suffer from the termination problem. Motivated by the above problems, we propose SimpleFT in this paper, which is a simple asynchronous consensus and is mainly inspired by the simplicity of the Bitcoin protocol. With a re-understanding of the Bitcoin protocol, we disassemble the life cycle of a block into three phases, namely proposal, dissemination, and confirmation. Corresponding to these phases, we devise or introduce the sortition algorithm, reliable broadcast algorithm, and quorum certificate mechanism in SimpleFT, respectively. To make full use of the network resources and improve the system throughput, we further introduce the layered architecture to SimpleFT, which enables multiple blocks to be confirmed at the same height. Comprehensive analysis is made to validate the correctness of SimpleFT and various experiments are conducted to demonstrate its efficient performance.

The transition to post-quantum cryptography has been an enormous challenge and effort for cryptographers over the last decade, with impressive results such as the future NIST standards. However, the latter has so far only considered central cryptographic mechanisms (signatures or KEM) and not more advanced ones, e.g., targeting privacy-preserving applications. Of particular interest is the family of solutions called blind signatures, group signatures and anonymous credentials, for which standards already exist, and which are deployed in billions of devices. Such a family does not have, at this stage, an efficient post-quantum counterpart although very recent works improved this state of affairs by offering two different alternatives: either one gets a system with rather large elements but a security proved under standard assumptions or one gets a more efficient system at the cost of ad-hoc interactive assumptions or weaker security models. Moreover, all these works have only considered size complexity without implementing the quite complex building blocks their systems are composed of. In other words, the practicality of such systems is still very hard to assess, which is a problem if one envisions a post-quantum transition for the corresponding systems/standards.

In this work, we propose a construction of so-called signature with efficient protocols (SEP), which is the core of such privacy-preserving solutions. By revisiting the approach by Jeudy et al. (Crypto 2023) we manage to get the best of the two alternatives mentioned above, namely short sizes with no compromise on security. To demonstrate this, we plug our SEP in an anonymous credential system, achieving credentials of less than 80 KB. In parallel, we fully implemented our system, and in particular the complex zero-knowledge framework of Lyubashevsky et al. (Crypto'22), which has, to our knowledge, not be done so far. Our work thus not only improves the state-of-the-art on privacy-preserving solutions, but also significantly improves the understanding of efficiency and implications for deployment in real-world systems.

While formal constructions for cryptographic schemes have steadily evolved and emerged over the past decades, the design and implementation of efficient and secure hardware instances is still a mostly manual, tedious, and intuition-driven process. With the increasing complexity of modern cryptography, e.g., Post-Quantum Cryptography (PQC) schemes, and consideration of physical implementation attacks, e.g., Side-Channel Analysis (SCA), the design space often grows exorbitantly without developers being able to weigh all design options.

This immediately raises the necessity for tool-assisted Design Space Exploration (DSE) for efficient and secure cryptographic hardware. For this, we present the progressive HADES framework, offering a customizable, extendable, and streamlined DSE for efficient and secure cryptographic hardware accelerators. This tool exhaustively traverses the design space driven by security requirements, rapidly predicts user-defined performance metrics, e.g., area footprint or cycle-accurate latency, and instantiates the most suitable candidate in a synthesizable Hardware Description Language (HDL).

We demonstrate the capabilities of our framework by applying our proof-of-concept implementation to a wide-range selection of state-of-the-art symmetric and PQC schemes, including the ChaCha20 stream cipher and the designated PQC standard Kyber, for which we provide the first set of arbitrary-order masked hardware implementations.

The prevailing ciphers rely on the weak assumption that their attacker is not smarter than expected by their designers. The resultant crypto ecology favors the cryptographic powerhouses, and hinders cyber freedom, cyber privacy and cyber democracy. This weakness can be remedied by using the gold standard of cryptography -- One Time Pad, OTP. Alas, it comes with a prohibitive cost of a key as long as the message it encrypts. When the stakes are high enough users pay this high price because OTP is immunized against smarter and better equipped attackers. Claude Shannon has shown that this size imposition on the key is non-negotiable in the context he analyzed. Alas, changing the context, one could achieve OTP equivalence. Three simple changes are introduced: (i) make the size of the key an integral part of the secret, (ii) every finite message is encrypted with an arbitrary part of the key, (iii) allow for open-ended dilution of the contents-bearing bits of the ciphertext, with content-devoid bits, which don't confuse the intended recipient, but impose an open-ended cryptanalytic barrier before the attacker. A-priori a cryptanalyst is facing a set of messages each of them deemed plausible to be the one hidden in the ciphertext. If the ciphertext is Finite Key OTP compliant then membership in this set will not change after an exhaustive cryptanalytic processing of the ciphertext. This constitutes functional equivalence with OTP. OTP functionality with a shared finite key creates a path to digital freedom, digital privacy and digital democracy.

We argue that there are some scenarios in which plausible deniability might be desired for a digital signature scheme. For instance, the non-repudiation property of conventional signature schemes is problematic in designing an Instant Messaging system (WPES 2004). In this paper, we formally define a non-binding signature scheme in which the Signer is able to disavow her own signature if she wants, but, the Verifier is not able to dispute a signature generated by the Signer. That is, the Signer is able to convince a third party Judge that she is the owner of a signature without disclosing her secret information. We propose a signature scheme that is non-binding and unforgeable. Our signature scheme is post-quantum secure if the underlying cryptographic primitives are post-quantum secure. In addition, a modification to our nonbinding signature scheme leads to an Instant Messaging system that is of independent interest.

A new security model for fully homomorphic encryption (FHE), called INDCPA-D security and introduced by Li and Micciancio [Eurocrypt'21], strengthens INDCPA-D security by giving the attacker access to a decryption oracle for ciphertexts for which it should know the underlying plaintexts. This includes ciphertexts that it (honestly) encrypted and those obtained from the latter by evaluating circuits that it chose. Li and Micciancio singled out the CKKS FHE scheme for approximate data [Asiacrypt'17] by giving an INDCPA-D attack on it and (erroneously) claiming that INDCPA-D security and INDCPA-D security coincide for FHEs on exact data.

We correct the widespread belief according to which INDCPA-D attacks are specific to approximate homomorphic computations. Indeed, the  equivalency formally proved by Li and Micciancio assumes that the schemes are not only exact but have a negligible probability of incorrect decryption. However, almost all competitive implementations of exact FHE schemes give away strong correctness by analyzing correctness heuristically and allowing noticeable probabilities of incorrect decryption. 

We exploit this imperfect correctness  to mount efficient indistinguishability and key-recovery attacks against all major exact FHE schemes.  We illustrate their strength by concretely breaking the default BFV implementation of OpenFHE and simulating an attack for the default parameter set of the CGGI implementation of TFHE-rs (the attack is too expensive to be run on commodity desktops, because of the cost of CGGI bootstrapping). Our attacks extend to threshold versions of the exact FHE schemes, when the correctness is similarly loose.

The search for differential characteristics on block ciphers is a difficult combinatorial problem. In this paper, we investigate the performances of an AI-originated technique, Single Player Monte-Carlo Tree Search (SP-MCTS), in finding good differential characteristics on ARX ciphers, with an application to the block cipher SPECK. In order to make this approach competitive, we include several heuristics, such as the combination of forward and backward searches, and achieve significantly faster results than state-of-the-art works that are not based on automatic solvers. We reach 9, 11, 13, 13 and 15 rounds for SPECK32, SPECK48, SPECK64, SPECK96 and SPECK128 respectively. In order to build our algorithm, we revisit Lipmaa and Moriai's algorithm for listing all optimal differential transitions through modular addition, and propose a variant to enumerate all transitions with probability close (up to a fixed threshold) to the optimal, while fixing a minor bug in the original algorithm.

A linear code is considered self-orthogonal if it is contained within its dual code. Self-orthogonal codes have applications in linear complementary dual codes, quantum codes and so on. The construction of self-orthogonal codes from functions over finite fields has been studied in the literature. In this paper, we generalize the construction method given by Heng et al. (2023) to weakly regular plateaued functions. We first construct several families of ternary self-orthogonal codes from weakly regular plateaued unbalanced functions. Then we use the self-orthogonal codes to construct new families of ternary LCD codes. As a consequence, we obtain (almost) optimal ternary self-orthogonal codes and LCD codes. Moreover, we propose two new classes of $p$-ary linear codes from weakly regular plateaued unbalanced functions over the finite fields of odd characteristics.

In conventional deep-learning-based side-channel attacks (DL-SCAs), an attacker trains a model by updating parameters to minimize the negative log-likelihood (NLL) loss function. Although a decrease in NLL improves DL-SCA performance, the reasons for this improvement remain unclear because of the lack of a formal analysis. To address this open problem, this paper explores the relationship between NLL and the attack success rate (SR) and conducts an information-theoretical analysis of DL-SCAs with an NLL loss function to solve open problems in DL-SCA. To this end, we introduce a communication channel for DL-SCAs and derive an inequality that links model outputs to the SR. Our inequality states that mutual information between the model output and intermediate value, which is named the latent perceived information (LPI), provides an upper bound of the SR of a DL-SCA with a trained neural network. Subsequently, we examine the conjecture by Ito et al. on the relationship between the effective perceived information (EPI) and SR and clarify its valid conditions from the perspective of LPI. Our analysis results reveal that a decrease in NLL correlated with an increase in LPI, which indicates that the model capability to extract intermediate value information from traces is enhanced. In addition, the results indicate that the LPI bounds the SR from above, and a higher upper bound of the SR could directly improve the SR if the selection function satisfies certain conditions, such as bijectivity. Finally, these theoretical insights are validated through attack experiments on neural network models applied to AES software and hardware implementations with masking countermeasures.

We study the complexity of memory checkers with computational security and prove the first general tight lower bound.

Memory checkers, first introduced over 30 years ago by Blum, Evans, Gemmel, Kannan, and Naor (FOCS '91, Algorithmica '94), allow a user to store and maintain a large memory on a remote and unreliable server by using small trusted local storage. The user can issue instructions to the server and after every instruction, obtain either the correct value or a failure (but not an incorrect answer) with high probability. The main complexity measure of interest is the size of the local storage and the number of queries the memory checker makes upon every logical instruction. The most efficient known construction has query complexity $O(\log n/\log\log n)$ and local space proportional to a computational security parameter, assuming one-way functions, where $n$ is the logical memory size. Dwork, Naor, Rothblum, and Vaikuntanathan (TCC '09) showed that for a restricted class of ``deterministic and non-adaptive'' memory checkers, this construction is optimal, up to constant factors. However, going beyond the small class of deterministic and non-adaptive constructions has remained a major open problem.

In this work, we fully resolve the complexity of memory checkers by showing that any construction with local space $p$ and query complexity $q$ must satisfy $$ p \ge \frac{n}{(\log n)^{O(q)}} \;. $$ This implies, as a special case, that $q\ge \Omega(\log n/\log\log n)$ in any scheme, assuming that $p\le n^{1-\varepsilon}$ for $\varepsilon>0$. The bound applies to any scheme with computational security, completeness $2/3$, and inverse polynomial in $n$ soundness (all of which make our lower bound only stronger). We further extend the lower bound to schemes where the read complexity $q_r$ and write complexity $q_w$ differ. For instance, we show the tight bound that if $q_r=O(1)$ and $p\le n^{1-\varepsilon}$ for $\varepsilon>0$, then $q_w\ge n^{\Omega(1)}$. This is the first lower bound, for any non-trivial class of constructions, showing a read-write query complexity trade-off.

Our proof is via a delicate compression argument showing that a ``too good to be true'' memory checker can be used to compress random bits of information. We draw inspiration from tools recently developed for lower bounds for relaxed locally decodable codes. However, our proof itself significantly departs from these works, necessitated by the differences between settings.

Payment channel networks are a promising solution to the scalability challenge of blockchains and are designed for significantly increased transaction throughput compared to the layer one blockchain. Since payment channel networks are essentially decentralized peer-to-peer networks, routing transactions is a fundamental challenge. Payment channel networks have some unique security and privacy requirements that make pathfinding challenging, for instance, network topology is not publicly known, and sender/receiver privacy should be preserved, in addition to providing atomicity guarantees for payments. In this paper, we present an efficient privacy-preserving routing protocol, SPRITE, for payment channel networks that supports concurrent transactions. By finding paths offline and processing transactions online, SPRITE can process transactions in just two rounds, which is more efficient compared to prior work. We evaluate SPRITE’s performance using Lightning Net- work data and prove its security using the Universal Composability framework. In contrast to the current cutting-edge methods that achieve rapid transactions, our approach significantly reduces the message complexity of the system by 3 orders of magnitude while maintaining similar latencies.

In its standard form, the AKS prime identification algorithm is deterministic and polynomial time but too slow to be of practical use. By dropping its deterministic attribute, it can be accelerated to an extent that it is practically useful, though still much slower than the widely used Miller-Rabin-Selfridge-Monier (MRSM) algorithm based on the Fermat Little Theorem or the Solovay-Strassen algorithm based on the Euler Criterion. The change made, in the last stage of AKS, is to check a modular equation not for a long sequence of values but for a single one. Another change is to reduce, arbitrarily, the size of the parameter $r$ giving the degree of the polynomial used for the last stage.

The Signal protocol and its X3DH key exchange core are regularly used by billions of people in applications like WhatsApp but are unfortunately not quantum-secure. Thus, designing an efficient and post-quantum secure X3DH alternative is paramount. Notably, X3DH supports asynchronicity, as parties can immediately derive keys after uploading them to a central server, and deniability, allowing parties to plausibly deny having completed key exchange. To satisfy these constraints, existing post-quantum X3DH proposals use ring signatures (or equivalently a form of designated-verifier signatures) to provide authentication without compromising deniability as regular signatures would. Existing ring signature schemes, however, have some drawbacks. Notably, they are not generally proven secure in the quantum random oracle model (QROM) and so the quantum security of parameters that are proposed is unclear and likely weaker than claimed. In addition, they are generally slower than standard primitives like KEMs.

In this work, we propose an efficient, deniable and post-quantum X3DH-like protocol that we call K-Waay, that does not rely on ring signatures. At its core, K-Waay uses a split-KEM, a primitive introduced by Brendel et al. [SAC 2020], to provide Diffie-Hellman-like implicit authentication and secrecy guarantees. Along the way, we revisit the formalism of Brendel et al. and identify that additional security properties are required to prove a split-KEM-based protocol secure. We instantiate split-KEM by building a protocol based on the Frodo key exchange protocol relying on the plain LWE assumption: our proofs might be of independent interest as we show it satisfies our novel unforgeability and deniability security notions. Finally, we complement our theoretical results by thoroughly benchmarking both K-Waay and existing X3DH protocols. Our results show even when using plain LWE and a conservative choice of parameters that K-Waay is significantly faster than previous work.

In recent breakthrough results, novel use of garbled circuits yielded constructions for several primitives like Identity-Based Encryption (IBE) and 2-round secure multi-party computation, based on standard assumptions in public-key cryptography. While the techniques in these different results have many common elements, these works did not offer a modular abstraction that could be used across them.

Our main contribution is to introduce a novel notion of obfuscation, called Reach-Restricted Reactive Program Obfuscation (R3PO) that captures the essence of these constructions, and exposes additional capabilities. We provide a powerful composition theorem whose proof fully encapsulates the use of garbled circuits in these works. As an illustration of the potential of R3PO, and as an important contribution of independent interest, we present a variant of Multi-Authority Attribute-Based Encryption (MA-ABE) that can be based on (single-authority) CP-ABE in a blackbox manner, using only standard cryptographic assumptions (e.g., DDH). This is in stark contrast to the existing constructions for MA-ABE, which rely on the random oracle model and/or support only limited policy classes.

Homomorphic encryption is a powerful privacy-preserving technology that is notoriously difficult to configure and use, even for experts. The key difficulties include restrictive programming models of homomorphic schemes and choosing suitable parameters for an application. In this tutorial, we outline methodologies to solve these issues and allow for conversion of any application to the encrypted domain using both leveled and fully homomorphic encryption.

The first approach, called Walrus, is suitable for arithmetic-intensive applications with limited depth and applications with high throughput requirements. Walrus provides an intuitive programming interface and handles parameterization automatically by analyzing the application and gathering statistics such as homomorphic noise growth to derive a parameter set tuned specifically for the application. We provide an in-depth example of this approach in the form of a neural network inference as well as guidelines for using Walrus effectively.

Conversely, the second approach (HELM) takes existing HDL designs and converts them to the encrypted domain for secure outsourcing on powerful cloud servers. Unlike Walrus, HELM supports FHE backends and is well-suited for complex applications. At a high level, HELM consumes netlists and is capable of performing logic gate operations homomorphically on encryptions of individual bits. HELM incorporates both CPU and GPU acceleration by taking advantage of the inherent parallelism provided by Boolean circuits. As a case study, we walk through the process of taking an off-the-shelf HDL design in the form of AES-128 decryption and running it in the encrypted domain with HELM.