International Association
for Cryptologic Research

Fully Homomorphic Encryption (FHE) has emerged as a promising technology for processing encrypted data without the need for decryption. Despite its potential, its practical implementation has faced challenges due to substantial computational overhead. To address this issue, we propose the $first$ chiplet-based FHE accelerator design `REED', which enables scalability and offers high throughput, thereby enhancing homomorphic encryption deployment in real-world scenarios. It incorporates well-known wafer yield issues during fabrication which significantly impacts production costs. In contrast to state-of-the-art approaches, we also address data exchange overhead by proposing a non-blocking inter-chiplet communication strategy. We incorporate novel pipelined Number Theoretic Transform and automorphism techniques, leveraging parallelism and providing high throughput.

Experimental results demonstrate that REED 2.5D integrated circuit consumes 177 mm$^2$ chip area, 82.5 W average power in 7nm technology, and achieves an impressive speedup of up to 5,982$\times$ compared to a CPU (24-core 2$\times$Intel X5690), and 2$\times$ better energy efficiency and 50\% lower development cost than state-of-the-art ASIC accelerator. To evaluate its practical impact, we are the $first$ to benchmark an encrypted deep neural network training. Overall, this work successfully enhances the practicality and deployability of fully homomorphic encryption in real-world scenarios.

Vehicle-to-grid (V2G) networks, as an emerging smart grid paradigm, can be integrated with renewable energy resources to provide power services and manage electricity demands. When accessing electricity services, an electric vehicle(EV) typically provides authentication or/and payment information containing identifying data to a service provider, which raises privacy concerns as malicious entities might trace EV activity or exploit personal information. Although numerous anonymous authentication and payment schemes have been presented for V2G networks, no such privacy-preserving scheme supports authentication and payment simultaneously. Therefore, this paper is the first to present a privacy-preserving authentication scheme with anonymous payment for V2G networks (PAP, for short). In addition, this scheme also supports accountability and revocability, which are practical features to prevent malicious behavior; minimal attribute disclosure, which maximizes the privacy of EV when responding to the service provider's flexible access policies; payment binding, which guarantees the accountability in the payment phase; user-controlled linkability, which enables EV to decide whether different authentication sessions are linkable for continuous services. On the performance side, we implement PAP with the pairing cryptography library, then evaluate it on different hardware platforms, showing that it is essential for V2G applications.

The KpqC competition has begun in 2022, that aims to standardize Post-Quantum Cryptography (PQC) in the Republic of Korea. Among the 16 submissions of the KpqC competition, the lattice-based schemes exhibit the most promising and balanced features in performance. In this paper, we propose an effective classical CCA attack to recover the transmitted session key for NTRU+, one of the lattice-based Key Encapsulation Mechanisms (KEM) proposed in the KpqC competition, for the first time. With the proposed attacks, we show that all the suggested parameters of NTRU+ do not satisfy the claimed security. We also suggest a way to modify the NTRU+ scheme to defend our attack.

In this paper we continue the study of two-round broadcast-optimal MPC, where broadcast is used in one of the two rounds, but not in both. We consider the realistic scenario where the round that does not use broadcast is asynchronous. Since a first asynchronous round (even when followed by a round of broadcast) does not admit any secure computation, we introduce a new notion of asynchrony which we call $(t_d, t_m)$-asynchrony. In this new notion of asynchrony, an adversary can delay or drop up to $t_d$ of a given party's incoming messages; we refer to $t_d$ as the deafness threshold. Similarly, the adversary can delay or drop up to $t_m$ of a given party's outgoing messages; we refer to $t_m$ as the muteness threshold.

We determine which notions of secure two-round computation are achievable when the first round is $(t_d, t_m)$-asynchronous, and the second round is over broadcast. Similarly, we determine which notions of secure two-round computation are achievable when the first round is over broadcast, and the second round is (fully) asynchronous. We consider the cases where a PKI is available, when only a CRS is available but private communication in the first round is possible, and the case when only a CRS is available and no private communication is possible before the parties have had a chance to exchange public keys.

The encryption processes and cryptosystems are very important. We use them to protect our private information over the Internet. Cellular automata are ones of the computational models that can also be used in cryptosystems. The advantage of the cellular automata is their abilities to work in parallel, and thus can reduce the encryption time. Some applications require the encryption time to be small, so this paper aims to reduce the encryption time of the cellular automata cryptosystems. We propose a new technique to permit the cryptosystems to get the avalanche effect faster. This avalanche effect is one of the desired properties for cryptosystems. In the proposed technique, the new type of neighbor is defined, a sequence of neighbor tuples. We apply our technique to Seredynski and Bouvry’s work, and the results show that the number of iterations can be reduced up to three times. This makes our cellular automata cryptosystems run faster. The relationship between the size of the neighbor and the size of the cellular automata, and the effect of neighbor sequences to the hardware implementations are left for further studies.

Zero-knowledge range proofs play a critical role in confidential transactions (CT) on blockchain systems. They are used to prove the non-negativity of committed transaction payments without disclosing the exact values. Logarithmic-sized range proofs with transparent setups, e.g., Bulletproofs, which aim to prove a committed value lies in the range $[0, 2^N-1]$ where $N$ is the bit length of the range, have gained growing popularity for communication-critical blockchain systems as they increase scalability by allowing a block to accommodate more transactions. In this paper, we propose SwiftRange, a new type of logarithmic-sized zero-knowledge range argument with a transparent setup in the discrete logarithm setting. Our argument can be a drop-in replacement for range proofs in blockchain-based confidential transactions. Compared with Bulletproofs, our argument has higher computational efficiency and lower round complexity while incurring comparable communication overheads for CT-friendly ranges, where $N \in \{32,64\}$. Specifically, a SwiftRange achieves 1.61$\times$ and 1.32$\times$ proving efficiency with no more than 1.1$\times$ communication costs for both ranges, respectively. More importantly, our argument offers a $2.3\times$ increase in verification efficiency. Furthermore, our argument has a smaller size when $N \leq 16$, making it competitive for many other communication-critical applications. Our argument supports the aggregation of multiple single arguments for greater efficiency in communication and verification. Finally, we benchmarked our argument against the state-of-the-art range proofs to demonstrate its practicality.

In this work, we propose a novel single-trace key recovery attack targeting side-channel leakage from the key-generation procedure of Kyber KEM. Our attack exploits the inherent nature of the Module-Learning With Errors (Module-LWE) problem used in Kyber KEM. We demonstrate that the inherent reliance of Kyber KEM on the Module-LWE problem results in a higher number of repeated computations with the secret key, compared to the Ring-LWE problem of similar security level. We exploit leakage from the pointwise multiplication operation in the key-generation procedure, and take advantage of the properties of the Module-LWE instance to enable a potential single trace key recovery attack. We validated the efficacy of our attack on both simulated and real traces, and we performed experiments on both the reference and assembly optimized implementation of Kyber KEM, taken from the pqm4 library, a well-known benchmarking and testing framework for PQC schemes on the ARM Cortex-M4 microcontroller. We also analyze the applicability of our attack on the countermeasures against traditional SCA such as masking and shuffling. We believe our work motivates more research towards SCA resistant implementation of key-generation procedure for Kyber KEM.

Time-Lock puzzles (TLP) are cryptographic protocols that enable a client to lock a message in such a way that a server can only unlock it after a specific time period. However, existing TLPs have certain limitations: (i) they assume that both the client and server always possess sufficient computational resources and (ii) they solely focus on the lower time bound for finding a solution, disregarding the upper bound that guarantees a regular server can find a solution within a certain time frame. Additionally, existing TLPs designed to handle multiple puzzles either (a) entail high verification costs or (b) lack generality, requiring identical time intervals between consecutive solutions. To address these limitations, this paper introduces, for the first time, the concept of a "Delegated Time-Lock Puzzle" and presents a protocol called "Efficient Delegated Time- Lock Puzzle" (ED-TLP) that realises this concept. ED-TLP allows the client and server to delegate their resource-demanding tasks to third-party helpers. It facilitates real-time verification of solution correctness and efficiently handles multiple puzzles with varying time intervals. ED-TLP ensures the delivery of solutions within predefined time limits by incorporating both an upper bound and a fair payment algorithm. We have implemented ED-TLP and conducted a comprehensive analysis of its overheads, demonstrating the efficiency of the construction

Side-channel attacks are a fundamental threat to the security of cryptographic implementations. One of the most prominent countermeasures against side-channel attacks is masking, where each intermediate value of the computation is secret shared, thereby concealing the computation's sensitive information. An important security model to study the security of masking schemes is the random probing model, in which the adversary obtains each intermediate value of the computation with some probability $p$. To construct secure masking schemes, an important building block is the refreshing gadget, which updates the randomness of the secret shared intermediate values. Recently, Dziembowski, Faust, and Zebrowski (ASIACRYPT'19) analyzed the security of a simple refreshing gadget by using a new technique called the leakage diagram.

In this work, we follow the approach of Dziembowski et al. and significantly improve its methodology. Concretely, we refine the notion of a leakage diagram via so-called dependency graphs, and show how to use this technique for arbitrary complex circuits via composition results and approximation techniques. To illustrate the power of our new techniques, as a case study, we designed provably secure parallel gadgets for the random probing model, and adapted the ISW multiplication such that all gadgets can be parallelized. Finally, we evaluate concrete security levels, and show how our new methodology can further improve the concrete security level of masking schemes. This results in a compiler provable secure up to a noise level of $ O({1})$ for affine circuits and $ O({1}/{\sqrt{n}})$ in general.

An attribute-based credential system enables users to prove possession of a credential and statements over certified attributes to verifiers in zero-knowledge while maintaining anonymity and unlinkability. In a relational anonymous credential system, users can further prove their relationship to other entities in their social graph, such as position in an organizational hierarchy or friends-of-friends status in an online social network graph, while protecting their own privacy and that of other users involved in the social graph. While traditional anonymous credential schemes make no provisions for privacy-preserving relationship predicates, a relational credential system is more usable, because it can facilitate relationship-based access control with a wide range of predicates and offers strong privacy guarantees for relationship proofs. We propose the first relational credential scheme, based on a new $q$-SDH graph signature scheme and an efficient zero-knowledge proof system for graph predicates. We rigorously prove the security for the proposed scheme and provide a benchmark using Facebook social graphs.

Blockchain has attracted significant attention in recent years due to its potential to revolutionize various industries by providing trustlessness. To comprehensively examine blockchain systems, this article presents both a macro-level overview on the most popular blockchain systems, and a micro-level analysis on a general blockchain framework and its crucial components. The macro-level exploration provides a big picture on the endeavors made by blockchain professionals over the years to enhance the blockchain performance while the micro-level investigation details the blockchain building blocks for deep technology comprehension. More specifically, this article introduces a general modular blockchain analytic framework that decomposes a blockchain system into interacting modules and then examines the major modules to cover the essential blockchain components of network, consensus, and distributed ledger at the micro-level. The framework as well as the modular analysis jointly build a foundation for designing scalable, flexible, and application-adaptive blockchains that can meet diverse requirements. Additionally, this article explores popular technologies that can be integrated with blockchain to expand functionality and highlights major challenges. Such a study provides critical insights to overcome the obstacles in designing novel blockchain systems and facilitates the further development of blockchain as a digital infrastructure to service new applications.

Side-channel attacks against cryptographic implementations are mitigated by the application of masking and hiding countermeasures. Hiding countermeasures attempt to reduce the Signal-to-Noise Ratio of measurements by adding noise or desynchronization effects during the execution of the cryptographic operations. To bypass these protections, attackers adopt signal processing techniques such as pattern alignment, filtering, averaging, or resampling. Convolutional neural networks have shown the ability to reduce the effect of countermeasures without the need for trace preprocessing, especially alignment, due to their shift invariant property. Data augmentation techniques are also considered to improve the regularization capacity of the network, which improves generalization and, consequently, reduces the attack complexity.

In this work, we deploy systematic experiments to investigate the benefits of data augmentation techniques against masked AES implementations when they are also protected with hiding countermeasures. Our results show that, for each countermeasure and dataset, a specific neural network architecture requires a particular data augmentation configuration to achieve significantly improved attack performance. Our results clearly show that data augmentation should be a standard process when targeting datasets with hiding countermeasures in deep learning-based side-channel attacks.

The open-source hardware IP model has recently started gaining popularity in the developer community. This model offers the integrated circuit (IC) developers wider standardization, faster time-to-market and richer platform for research. In addition, open-source hardware conforms to the Kerckhoff’s principle of a publicly-known algorithm and thus helps to enhance security. However, when security comes into consideration, source transparency is only one part of the solution. A complex global IC supply chain stands between the source and the final product. Hence, even if the source is known, the finished product is not guaranteed to match it. In this article, we propose the Open Scan model, in which, in addition to the source code, the IC vendor contributes a library-independent information on scan insertion. With scan information available, the user or a certification lab can perform partial reverse engineering of the IC to verify conformance to the advertised source. Compliance lists of open-source programs, such as of the OpenTitan cryptographic IC, can be amended to include this requirement. The Open Scan model addresses accidental and dishonest deviations from the golden model and partially addresses malicious modifications, known as hardware Trojans. We verify the efficiency of the proposed method in simulation with the Trust-Hub Trojan benchmarks and with several open-source benchmarks, in which we randomly insert modifications.

A linkable ring signature allows a user to sign anonymously on behalf of a group while ensuring that multiple signatures from the same user are detected. Applications such as privacy-preserving e-voting and e-cash can leverage linkable ring signatures to significantly improve privacy and anonymity guarantees. To scale to systems involving large numbers of users, short signatures with fast verification are a must. Concretely efficient ring signatures currently rely on a trusted authority maintaining a master secret, or follow an accumulator-based approach that requires a trusted setup.

In this work, we construct the first linkable ring signature with both logarithmic signature size and verification that does not require any trusted mechanism. Our scheme, which relies on discrete-log type assumptions and bilinear maps, improves upon a recent concise ring signature called DualRing by integrating improved preprocessing arguments to reduce the verification time from linear to logarithmic in the size of the ring. Our ring signature allows signatures to be linked based on what message is signed, ranging from linking signatures on any message to only signatures on the same message.

We provide benchmarks for our scheme and prove its security under standard assumptions. The proposed linkable ring signature is particularly relevant to use cases that require privacy-preserving enforcement of threshold policies in a fully decentralized context, and e-voting.

Oblivious Pseudo-Random Functions (OPRFs) are a central tool for building modern protocols for authentication and distributed computation. For example, OPRFs enable simple login protocols that do not reveal the password to the provider, which helps to mitigate known shortcomings of password-based authentication such as password reuse or mix-up. Reliable treatment of passwords becomes more and more important as we login to a multitude of services with different passwords in our daily life. To ensure the security and privacy of such services in the long term, modern protocols should always consider the possibility of attackers with quantum computers. Therefore, recent research has focused on constructing post-quantum-secure OPRFs. Unfortunately, existing constructions either lack efficiency, or they are based on complex and relatively new cryptographic assumptions, some of which have lately been disproved. In this paper, we revisit the security and the efficiency of the well-known “OPRFs via Garbled Circuits” approach. Such an OPRF is presumably post-quantum-secure and built from well-understood primitives, namely symmetric cryptography and oblivious transfer. We investigate security in the strong Universal Composability model, which guarantees security even when multiple instances are executed in parallel and in conjunction with arbitrary other protocols, which is a realistic scenario in today’s internet. At the same time, it is faster than other current post-quantumsecure OPRFs. Our implementation and benchmarks demonstrate that our proposed OPRF is currently among the best choices if the privacy of the data has to be guaranteed for a long time.

We present new protocols for threshold Schnorr signatures that work in an asynchronous communication setting, providing robustness and optimal resilience. These protocols provide unprecedented performance in terms of communication and computational complexity. In terms of communication complexity, for each signature, a single party must transmit a few dozen group elements and scalars across the network (independent of the size of the signing committee). In terms of computational complexity, the amortized cost for one party to generate a signature is actually less than that of just running the standard Schnorr signing or verification algorithm (at least for moderately sized signing committees, say, up to 100).

For example, we estimate that with a signing committee of 49 parties, at most 16 of which are corrupt, we can generate 50,000 Schnorr signatures per second (assuming each party can dedicate one standard CPU core and 500Mbs of network bandwidth to signing). Importantly, this estimate includes both the cost of an offline precomputation phase (which just churns out message independent "presignatures") and an online signature generation phase. Also, the online signing phase can generate a signature with very little network latency (just one to three rounds, depending on how throughput and latency are balanced).

To achieve this result, we provide two new innovations. One is a new secret sharing protocol (again, asynchronous, robust, optimally resilient) that allows the dealer to securely distribute shares of a large batch of ephemeral secret keys, and to publish the corresponding ephemeral public keys. To achieve better performance, our protocol minimizes public-key operations, and in particular, is based on a novel technique that does not use the traditional technique based on "polynomial commitments". The second innovation is a new algorithm to efficiently combine ephemeral public keys contributed by different parties (some possibly corrupt) into a smaller number of secure ephemeral public keys. This new algorithm is based on a novel construction of a so-called "super-invertible matrix" along with a corresponding highly-efficient algorithm for multiplying this matrix by a vector of group elements.

As protocols for verifiably sharing a secret key with an associated public key and the technology of super-invertible matrices both play a major role in threshold cryptography and multi-party computation, our two new innovations should have applicability well beyond that of threshold Schnorr signatures.

The recent advancements in deep learning have brought about significant changes in various aspects of people's lives. Meanwhile, these rapid developments have raised concerns about the legitimacy of the training process of deep networks. However, to protect the intellectual properties of untrusted AI developers, directly examining the training process by accessing the model parameters and training data by verifiers is often prohibited.

In response to this challenge, we present zkDL, an efficient zero-knowledge proof of deep learning training. At the core of zkDL is zkReLU, a specialized zero-knowledge proof protocol with optimized proving time and proof size for the ReLU activation function, a major obstacle in verifiable training due to its non-arithmetic nature. To integrate zkReLU into the proof system for the entire training process, we devise a novel construction of an arithmetic circuit from neural networks. By leveraging the abundant parallel computation resources, this construction reduces proving time and proof sizes by a factor of the network depth. As a result, zkDL enables the generation of complete and sound proofs, taking less than a minute with a size of less than 20 kB per training step, for a 16-layer neural network with 200M parameters, while ensuring the privacy of data and model parameters.

We give the first construction of a fully black-box round-optimal secure multiparty computation (MPC) protocol in the plain model. Our protocol makes black-box use of a sub-exponentially secure two-message statistical sender private oblivious transfer (SSP-OT), which in turn can be based on (sub-exponential variants of) almost all of the standard cryptographic assumptions known to imply public-key cryptography.

This work focuses on the parallel broadcast primitive, where each of the $n$ parties wish to broadcast their $\ell$-bit input in parallel. We consider the authenticated model with PKI and digital signatures that is secure against $t < n/2$ Byzantine faults under a synchronous network.

We show a generic reduction from parallel broadcast to a new primitive called graded parallel broadcast and a single instance of validated Byzantine agreement. Using our reduction, we obtain parallel broadcast protocols with $O(n^2 \ell + \kappa n^3)$ communication ($\kappa$ denotes a security parameter) and expected constant rounds. Thus, for inputs of size $\ell = \Omega(n)$ bits, our protocols are asymptotically free.

Our graded parallel broadcast uses a novel gradecast protocol with multiple grades with asymptotically optimal communication complexity of $O(n \ell + \kappa n^2)$ for inputs of size $\ell$ bits. We also present a multi-valued validated Byzantine agreement protocol with asymptotically optimal communication complexity of $O(n \ell + \kappa n^2)$ for inputs of size $\ell$ bits in expectation and expected constant rounds. Both of these primitives are of independent interest.

Byzantine fault-tolerant state machine replication (BFT-SMR) replicates a state machine across a set of replicas, and processes requests as a single machine even in the presence of Byzantine faults. Recently, synchronous BFT-SMRs have received tremendous attention due to their simple design and high fault-tolerance threshold.

In this paper, we propose Arena, the first multi-leader synchronous BFT-SMR. Thanks to the synchrony assumption, Arena gains the performance benefit from multi-leader with a much simpler design (compared to other partially synchronous multi-leader designs). Furthermore, it is more robust: ``no progress'' of a leader will not trigger a view-change. Our experimental results show that Arena achieves a peak throughput of up to 7.7$\times$ higher than the state-of-the-art.