International Association
for Cryptologic Research

This paper proposes Oryx, a system for efficiently detecting cycles in federated graphs where parts of the graph are held by different parties and are private. Cycle detection is an important building block in designing fraud detection algorithms that operate on confidential transaction data held by different financial institutions. Oryx allows detecting cycles of various length while keeping the topology of the graphs secret, and it does so efficiently; Oryx achieves quasilinear computational complexity and scales well with more machines thanks to a parallel design. Our implementation of Oryx running on a single 32-core AWS machine (for each party) can detect cycles of up to length 6 in under 5 hours in a financial transaction graph that consists of tens of millions of nodes and edges. While the costs are high, adding more machines further reduces the completion time. Furthermore, Oryx is, to our knowledge, the first and only system that can handle this task.

Oblivious transfer (OT) is a fundamental cryptographic protocol that plays a crucial role in secure multi-party computation (MPC). Most practical OT protocols by, e.g., Naor and Pinkas (SODA'01) or Chou and Orlandi (Latincrypt'15), are based on Diffie-Hellman (DH)-like assumptions and not post-quantum secure. In contrast, many other components of MPC protocols, including garbled circuits and secret sharings, are post-quantum secure. The reliance on non-post-quantum OT protocols presents a significant security bottleneck with the advent of quantum computing. In this paper, we address this issue by constructing a simple, efficient OT protocol based on Saber, a Mod-LWR-based key exchange protocol. We implemented our OT protocol and conducted experiments to evaluate its performance. Our results show that our OT protocol significantly outperforms the state-of-the-art Kyber-based post-quantum OT protocol by Masny and Rindal (CCS'19) in terms of both computation and communication costs. Furthermore, the computation speed of our OT protocol is faster than the best-known DH-based OT protocol by Chou and Orlandi (Latincrypt'15), making it competitive to replace DH-based OT in the high-bandwidth network setting.

This paper reports the experimental results related to lineage event storage via smart contracts deployed on private and public blockchain. In our experiments we measure the following three metrics: the cost to deploy the storage smart contract on the blockchain, which measures the initial expenditure, typically in gas units, required to deploy the smart contract that facilitates lineage event storage, then the time and gas costs needed to store a lineage event. We investigated both single and multi-clients scenarios. We considered the following public blockchains: Hedera, Fantom, Harmony Shard0, Polygon Amoy, Ethereum Sepolia, Optimism Sepolia, Klaytn Baobab and Arbitrum Sepolia. Furthermore, we investigate the performances of Hyperledger Besu with different consensus algorithms as private blockchains.

We propose time-memory trade-off algorithms for evaluating look-up table (LUT) in both the leveled homomorphic encryption (LHE) and fully homomorphic encryption (FHE) modes in TFHE. For an arbitrary $n$-bit Boolean function, we reduce evaluation time by a factor of $O(n)$ at the expense of an additional memory of "only" $O(2^n)$ as a trade-off: The total asymptotic memory is also $O(2^n)$, which is the same as that of prior works. Our empirical results demonstrate that a $7.8 \times$ speedup in runtime is obtained with a $3.8 \times$ increase in memory usage for 16-bit Boolean functions in the LHE mode. Additionally, in the FHE mode, we achieve reductions in both runtime and memory usage by factors of $17.9 \times$ and $2.5 \times $, respectively, for 8-bit Boolean functions. The core idea is to decompose the function $f$ into sufficiently small subfunctions and leverage the precomputed results for these subfunctions, thereby achieving significant performance improvements at the cost of additional memory.

A threshold signature scheme splits the signing key among $\ell$ parties, such that any $t$-subset of parties can jointly generate signatures on a given message. Designing concretely efficient post-quantum threshold signatures is a pressing question, as evidenced by NIST's recent call.

In this work, we propose, implement, and evaluate a lattice-based threshold signature scheme, Ringtail, which is the first to achieve a combination of desirable properties: (i) The signing protocol consists of only two rounds, where the first round is message-independent and can thus be preprocessed offline. (ii) The scheme is concretely efficient and scalable to $t \leq 1024$ parties. For $128$-bit security and $t = 1024$ parties, we achieve $13.4$ KB signature size and $10.5$ KB of online communication. (iii) The security is based on the standard learning with errors (LWE) assumption in the random oracle model. This improves upon the state-of-the-art (with comparable efficiency) which either has a three-round signing protocol [Eurocrypt'24] or relies on a new non-standard assumption [Crypto'24].

To substantiate the practicality of our scheme, we conduct the first WAN experiment deploying a lattice-based threshold signature, across 8 countries in 5 continents. We observe that an overwhelming majority of the end-to-end latency is consumed by network latency, underscoring the need for round-optimized schemes.

In this work we present $\mathsf{HERatio}$, a homomorphic encryption scheme that builds on the scheme of Brakerski, and Fan and Vercauteren. Our scheme naturally accepts Laurent polynomials as inputs, allowing it to work with rationals via their bounded base-$b$ expansions. This eliminates the need for a specialized encoder and streamlines encryption, while maintaining comparable efficiency to BFV. To achieve this, we introduce a new variant of the Polynomial Learning With Errors (PLWE) problem which employs Laurent polynomials instead of the usual ``classic'' polynomials, and provide a reduction to the PLWE problem.

Advanced Encryption Standard Galois/Counter Mode (AES-GCM) is the most widely used Authenticated Encryption with Associated Data (AEAD) algorithm in the world. In this paper, we analyze the use of GCM with all the Initialization Vector (IV) constructions and lengths approved by NIST SP 800-38D when encrypting multiple plaintexts with the same key. We derive attack complexities in both ciphertext-only and known-plaintext models, with or without nonce hiding, for collision attacks compromising integrity and confidentiality. Our analysis shows that GCM with random IVs provides less than 128 bits of security. When 96-bit IVs are used, as recommended by NIST, the security drops to less than 97 bits. Therefore, we strongly recommend NIST to forbid the use of GCM with 96-bit random nonces.

This work describes vulnerabilities in the specification of the AEAD packets as introduced in the novel LibrePGP specification that is implemented by the widely used GnuPG application and the AES-based AEAD schemes as well as the Key Wrap Algorithm specified in the Cryptographic Message Syntax (CMS). These new attacks exploit the possibility to downgrade AEAD or AES Key Wrap ciphertexts to valid legacy CFB- or CBC-encrypted related ciphertexts and require that the attacker learns the content of the legacy decryption result. This can happen either due to the human recipient returning the decryption output, which has entirely pseudo-random appearance, to the attacker or due to a programmatic decryption oracle in the receiving system. The attacks effect the decryption of low-entropy plaintext blocks in AEAD ciphertexts and, in the case of LibrePGP, also the manipulation of existing AEAD ciphertexts. For AES Key Wrap in CMS, full key decryption is possible. Some of the attacks require multiple successful oracle queries. The attacks thus demonstrate that CCA2 security is not achieved by the LibrePGP and CMS AEAD or Key Wrap encryption in the presence of a legacy cipher mode decryption oracle. The proper countermeasure to thwart the attacks is a key derivation that ensures the use of unrelated block cipher keys for the different encryption modes.

Online ride-sharing services (RSS) have become very popular owing to increased awareness of environmental concerns and as a response to increased traffic congestion. To request a ride, users submit their locations and route information for ride matching to a service provider (SP), leading to possible privacy concerns caused by leakage of users' location data. We propose QuickPool, an efficient SP-aided RSS solution that can obliviously match multiple riders and drivers simultaneously, without involving any other auxiliary server. End-users, namely, riders and drivers share their route information with SP as encryptions of the ordered set of points-of-interest (PoI) of their route from their start to end locations. SP performs a zone based oblivious matching of drivers and riders, based on partial route overlap as well as proximity of start and end points. QuickPool is in the semi-honest setting, and makes use of secure multi-party computation. We provide security proof of our protocol, perform extensive testing of our implementation and show that our protocol simultaneously matches multiple drivers and riders very efficiently. We compare the performance of QuickPool with state-of-the-art works and observe a speed up of 1.6 - 2$\times$.

Blockchain consensus, a.k.a. BFT SMR, are protocols enabling $n$ processes to decide on an ever-growing chain. The fastest known asynchronous one is called 2-chain VABA (PODC'21 and FC'22), and is used as fallback chain in Abraxas* (CCS'23). It has a claimed $9.5\delta$ expected latency when used for a single shot instance, a.k.a. an MVBA. We exhibit attacks breaking it. Hence, the title of the fastest asynchronous MVBA with quadratic messages complexity goes to sMVBA (CCS'22), with $10\delta$ expected latency. Our positive contributions are two new and complementary designs.

$\bullet$ 2PAC (2-phase asynchronous consensus). It has a simpler and lighter chaining than in previous approaches. Instantiated with either quadratic or cubic phases of voting, it yields:

2PAC$^\text{lean}$: $+90\%$ throughput and $9.5\delta$ expected latency, with quadratic ($O(n^2)$) messages complexity. In both 2-chain VABA and sMVBA (as if chained, with pipelining), the quorum-certified transactions which were produced in the worst-case 1/3 of views with a slow leader were dumped, so the work was lost. The simpler design of 2PAC inserts such blocks in straight-line in the chain. Thus, contrary to naive uncle-referencing, this comes with no computational overhead, yielding a net $+50\%$ throughput gain over chained sMVBA. Both the remaining throughput and latency ($-0.5\delta$) gains, come from the lighter interactive construction of proofs of consistency appended to proposed blocks, compared to sMVBA.

2PAC$^\text{BIG}$: the fastest asynchronous blockchain consensus with cubic ($O(n^3)$) messages complexity. Fault-free single shot MVBA runs decide in just $4\delta$, as soon as no message is delivered more than twice faster than others: GradedDAG (SRDS'23) required furthermore no messages reordering.

$\bullet$ Super Fast Pipelined Blocks. This is an upgrade of previous approaches for pipelining: in 2-chain VABA, Cordial Miners (DISC'23) and GradedDAG, a block pipelined by a leader in the middle of the view had almost twice larger latency than the non-pipelined block. Our design provides a fast path deciding the pipelined block with even smaller latency than the non-pipelined block. The fast delay is guaranteed in all executions with a fair scheduler, but remarkably, whatever the behaviors of faulty processes. Consistency is preserved by a lightweight mechanism, of one threshold signature appended per proposal. Instantiated with the previous protocols, it yields: s2PAC$^\text{lean}$, with fast decision of pipelined blocks in $4\delta$; s2PAC$^\text{BIG}$, in $3\delta$; and sGradedDAG, in $3\delta$.

Measuring the fluctuations of the clock phase of a target was identified as a leakage source on early electromagnetic side-channel investigations. Despite this, only recently was directly measuring the clock phase (or jitter) of digital signals from a target connected to being a source of exploitable leakage. As the phase of a clock output will be related to signal propagation delay through the target, and this propagation delay is related to voltage, this means that most digital devices perform an unintended phase modulation (PM) of their internal voltage onto clock output phases.

This paper first demonstrates an unprofiled CPA attack against a Cortex-M microcontroller using the phase of a clock output, observing the signal on both optically isolated and capacitively isolated paths. The unprofiled attack takes only 2-4x more traces than an attack using a classic shunt-resistor measurement.

It is then demonstrated how the JTAG bypass mode can be used to force a clock through a digital device. This forced clock signal can then be used as a highly effective oscilloscope that is located on the target device. As the attack does not require modifications to the device (such as capacitor removal or heat spreader removal) it is difficult to detect using existing countermeasures. The example attack over JTAG uses an unprofiled CPA attack, requiring only about 5x more traces than an ideal shunt-resistor based measurement. In addition, a version of this attack using a fault correlation analysis attack is also demonstrated.

Countermeasures are discussed, and a simple resampling countermeasure is tested. All tools both offensive and defensive presented in the paper have been released under open-source licenses.

Anticipating the advent of large quantum computers, NIST started a worldwide competition in 2016 aiming to define the next cryptographic standards. HQC is one of these post-quantum schemes still in contention, with four others already in the process of being standardized. In 2022, Guo et al. introduced a timing attack that exploited an inconsistency in HQC rejection sampling function to recover its secret key in 866,000 calls to an oracle. The authors of HQC updated its specification by applying an algorithm to sample vectors in constant time. A masked implementation of this function was then proposed for BIKE but it is not directly applicable to HQC. In this paper we propose a masked specification-compliant version of HQC vector sampling function which relies, to our knowledge, on the first masked implementation of the Barrett reduction.

We have proposed a novel FHE scheme that uniquely encodes the plaintext with noise in a way that prevents the increasing noise from overflowing and corrupting the plaintext. This allows users to perform computations on encrypted data smoothly. The scheme is constructed using the Chinese Remainder Theorem (CRT), supporting a predefined number of modular operations on encrypted plaintext without the need for bootstrapping. Although FHE recently became popular after Gentry's work and various developments have occurred in the last decade, the idea of "Fully Homomorphic Encryption (FHE)" scheme was first introduced in the 1970s by Rivest. The Chinese Remainder Theorem is one of the most suitable tools for developing a FHE Scheme because it forms a ring homomorphism \( Z_{p_1} \times Z_{p_2} \times \ldots \times Z_{p_k} \cong Z_{p_1 p_2 \ldots p_k} \). Various attempts have been made to develop a FHE using CRT, but most of them were unsuccessful, mainly due to the chosen plaintext attack (CPA). The proposed scheme overcomes the chosen plaintext attack. The scheme also adds random errors to the message during encryption. However, these errors are added in such a way that, when homomorphic operations are performed over encrypted data, the increasing values of errors never overwrite the values of the messages, as happens in LWE-based homomorphic schemes. Therefore, one can perform a predefined number of homomorphic operations (both addition and multiplication) without worrying about the increasing values of errors.

We address the black-box complexity of constructing pseudorandom functions (PRF) from pseudorandom generators (PRG). The celebrated GGM construction of Goldreich, Goldwasser, and Micali (Crypto 1984) provides such a construction, which (even when combined with Levin's domain-extension trick) has super-logarithmic depth. Despite many years and much effort, this remains essentially the best construction we have to date. On the negative side, one step is provided by the work of Miles and Viola (TCC 2011), which shows that a black-box construction which just calls the PRG once and outputs one of its output bits, cannot be a PRF.

In this work, we make significant further progress: we rule out black-box constructions of PRF from PRG that follow certain structural constraints, but may call the PRG adaptively polynomially many times. In particular, we define ``tree constructions" which generalize the GGM structure: they apply the PRG $G$ along a tree path, but allow for different choices of functions to compute the children of a node on the tree and to compute the next node on the computation path down the tree. We prove that a tree construction of logarithmic depth cannot be a PRF (while GGM is a tree construction of super-logarithmic depth). We also show several other results and discuss the special case of one-call constructions.

Our main results in fact rule out even weak PRF constructions with one output bit. We use the oracle separation methodology introduced by Gertner, Malkin, and Reingold (FOCS 2001), and show that for any candidate black-box construction $F^G$ from $G$, there exists an oracle relative to which $G$ is a PRG, but $F^G$ is not a PRF.

The multilinear extension of an $m$-variate function $f : \{0,1\}^m \to \mathbb{F}$, relative to a finite field $\mathbb{F}$, is the unique multilinear polynomial $\hat{f} : \mathbb{F}^m \to \mathbb{F}$ that agrees with $f$ on inputs in $\{0,1\}^m$. In this note we show how, given oracle access to $f : \{0,1\}^m \to \mathbb{F}$ and a point $z \in \mathbb{F}^m$, to compute $\hat{f}(z)$ using $O(2^m)$ field operations and only $O(m)$ space. This improves on a previous algorithm due to Vu et al. (SP, 2013), which similarly uses $O(2^m)$ field operations but requires $O(2^m)$ space. Furthermore, the number of field additions in our algorithm is about half of that in Vu et al.'s algorithm, whereas the number of multiplications is the same up to small additive terms.

We show that the scalar product protocol [IEEE Trans. Parallel Distrib. Syst. 2023, 1060-1066] is insecure against semi-honest server attack, not as claimed. Besides, its complexity increases exponentially with the number $n$, which cannot be put into practice.

In this paper we introduce new concept of tropical increasing matrices and then prove that two tropical increasing matrices are commute. Using this property, we modified Stickel’s protocol. This idea similar to [5] where modified Stickel’s protocol using commuting matrices (Linde De La Puente Matrices).

A Blind Signature Scheme (BSS) is a cryptographic primitive that enables a user to obtain a digital signature on a message from a signer without revealing the message itself. The standard security notion against malicious users for a BSS is One-More Unforgeability (OMUF). One of the earliest and most well-studied blind signature schemes is the Schnorr BSS, although recent results show it does not satisfy OMUF. On the other hand, the Schnorr BSS does satisfy the weaker notion of sequential OMUF --- which restricts adversaries to opening signing sessions one at a time --- in the Algebraic Group Model (AGM) + Random Oracle Model (ROM). In light of this result, a natural question arises: does the Schnorr BSS satisfy OMUF with regard to adversaries that open no more than a small number of signing sessions concurrently?

This paper serves as a first step towards characterizing the security of the Schnorr BSS in the limited concurrency setting. Specifically, we demonstrate that the Schnorr BSS satisfies OMUF when at most two signing sessions can be open concurrently (in the AGM+ROM). Our argument suggests that it is plausible that the Schnorr BSS satisfies OMUF for up to polylogarithmically many concurrent signing sessions.

With concerns about data privacy growing in a connected world, cryptography researchers have focused on fully homomorphic encryption (FHE) for promising machine learning as a service solutions. Recent advancements have lowered the computational cost by several orders of magnitude, but the latency of fully homomorphic neural networks remains a barrier to adoption. This work proposes using multi-exit neural networks (MENNs) to accelerate the FHE inference. MENNs are network architectures that provide several exit points along the depth of the network. This approach allows users to employ results from any exit and terminate the computation early, saving both time and power. First, this work weighs the latency, communication, accuracy, and computational resource benefits of running FHE-based MENN inference. Then, we present the TorMENNt attack that can exploit the user's early termination decision to launch a concrete side-channel on MENNs. We demonstrate that the TorMENNt attack can predict the private image classification output of an image set for both FHE and plaintext threat models. We discuss possible countermeasures to mitigate the attack and examine their effectiveness. Finally, we tie the privacy risks with a cost-benefit analysis to obtain a practical roadmap for FHE-based MENN adoption.

(Receiver) Anamorphic encryption, introduced by Persiano $ \textit{et al.}$ at Eurocrypt 2022, considers the question of achieving private communication in a world where secret decryption keys are under the control of a dictator. The challenge here is to be able to establish a secret communication channel to exchange covert (i.e. anamorphic) messages on top of some already deployed public key encryption scheme.

Over the last few years several works addressed this challenge by showing new constructions, refined notions and extensions. Most of these constructions, however, are either ad hoc, in the sense that they build upon specific properties of the underlying PKE, or impose severe restrictions on the size of the underlying anamorphic message space.

In this paper we consider the question of whether it is possible to have realizations of the primitive that are both generic and allow for large anamorphic message spaces. We give strong indications that, unfortunately, this is not the case.

Our first result shows that $ \textit{any black-box realization} $ of the primitive, i.e. any realization that accesses the underlying PKE only via oracle calls, $ \textit{must} $ have an anamorphic message space of size at most $poly(\lambda)$ ($\lambda$ security parameter).

Even worse, if one aims at stronger variants of the primitive (and, specifically, the notion of asymmetric anamorphic encryption, recently proposed by Catalano $ \textit{et al.} $) we show that such black-box realizations are plainly impossible, i.e. no matter how small the anamorphic message space is.

Finally, we show that our impossibility results are rather tight: indeed, by making more specific assumptions on the underlying PKE, it becomes possible to build generic AE where the anamorphic message space is of size $\Omega(2^\lambda)$.