Zero-Knowledge Proof (ZKP) has been widely applied in fields such as blockchain and privacy-preserving computing. However, the proof generation process remains computationally complex and time-consuming, which limits its further applications. Various schemes have been proposed to optimize the underlying modular operations with dedicated hardware support, but existing schemes still face low-efficiency problems. To address the problems, we propose an efficient Barrett modular multiplier design, especially for ZKP. Evaluation on a Xilinx XCVU9P FPGA shows that, compared to two existing pipelined designs, the proposed design improves throughput per slice by up to 20.4% and 49.6%, respectively, and achieves an 8.6x improvement over an existing non-pipelined design.
The flexibility, scalability, and robustness of Unmanned Aerial Vehicle (UAV) swarms make them highly adaptable to dynamic environments. However, these advantages also introduce a range of optimization challenges, including constraints on size, weight, and energy consumption, as well as less explored topics such as lightweight, secure communication and reliable mission completion. Inspired by the ant identification methods, we propose a decentralized solution to enhance communication security and ensure mission completion within UAV swarms. Unlike existing methods that rely on Physical Unclonable Functions (PUFs), our solution introduces a PUF-less framework. Specifically, we define private IDs and group keys for individual UAVs, enabling secure communication while minimizing the risk of intrusion. The effectiveness of our method is validated through simulations, which highlight the importance of periodically updating the group key. Additionally, we address targeted attacks on private IDs by incorporating behavioral analysis and shared monitoring among swarm members. Our proposed approach paves the way for realizing the full potential of UAV swarms.
With the standardization of the lattice-based cryptographic scheme Kyber, physical security concerns in its hardware implementations, particularly their vulnerability to side-channel analysis (SCA), has emerged as a core challenge. Existing studies have shown that the polynomial multiplication (PWM) module, characterized by deterministic computation patterns and direct dependency on secret keys, constitutes a primary attack surface. To address this threat, this paper proposes DASH-T (Dynamic Address Shuffling and Temporal-hiding), a novel microarchitectural hybrid countermeasure integrating two defense mechanisms across distinct dimensions: (1) fine-grained address shuffling within PWM operations, leveraging a configurable permutation table and unique address remapping logic to completely obfuscate the physical execution sequence of sensitive operations; (2) a multi-dimensional random temporal hiding mechanism, triggered at random intervals, injecting delays of random durations. Evaluated on a FPGA platform, DASH-T effectively thwarts first-order correlation power analysis (CPA) attacks. Experimental results demonstrate that it increases the number of power traces required for successful key recovery by more than an order of magnitude in FPGA measurements, while suppressing t-values from test vector leakage assessment (TVLA) tests within the secure threshold of +/- 4.5. This robust security is achieved with only manageable overheads in hardware resources and performance. This work presents an efficient and practical countermeasure for the physical security design of post-quantum cryptographic hardware accelerators.
Paillier and ElGamal are the two most commonly used partial homomorphic encryption algorithms. However, they suffer from the computational bottleneck commonly faced by most HE schemes. To accelerate them, many accelerators have been proposed, but the existing accelerators only support one algorithm and do not have side-channel protection. Therefore, this paper proposes a unified hardware accelerator for Paillier and ElGamal, named HAPE, which is also able to resist side-channel attacks. In HAPE, in order to alleviate side-channel leakage, Montgomery modular multiplication without conditional subtraction is adopted, and an improved randomized Montgomery ladder algorithm is proposed for modular exponentiation. In addition, HAPE also integrates both modulus masking and exponential masking to further protect HAPE. Experimental results show that the sidechannel attack resistance of HAPE is significantly improved at the cost of 11.29% area increase, 35.70% power increase and 3.14% performance degradation. Compared with other related works, HAPE not only achieves a performance improvement of 2.51x to 75.61x, but also supports both Paillier and ElGamal algorithms, and has side-channel attack protection.
Model-Based Diagnosis (MBD), a prominent technique for identifying the causes of system malfunctions, is an important research area in reliability engineering. Existing MBD methods generate a large number of candidate diagnoses for fault localization, which requires extensive computational cost. This work introduces a novel fault propagation model to replace the model used in conventional MBD approaches, and proposes a novel diagnosis method upon this model to improve diagnosis performance. First, we generate fault propagation constraints to capture the logical relationships as faults propagate between two connected components, which can be used to narrow down some subsets of candidate diagnoses in the subsequent diagnosis phase. Second, we formally model fault propagation through Weighted Conjunctive Normal Form (WCNF), encoding it as a Maximum Satisfiability (MaxSAT) problem. Building on this model, we use RC2-an efficient MaxSAT solver-to compute only a subset of candidate diagnoses, thereby significantly reducing the computational cost of fault localization. Experimental evaluations using ISCAS-85 and ITC-99 benchmarks demonstrate that FPMBD achieves a reduction in the number of diagnoses typically ranging from 1 to 5 orders of magnitude, along with a reduction in diagnosis time ranging from 1 to 4 orders of magnitude in almost all of the test cases, compared to state-of-theart algorithms, including CMMO, HSD, One-SAT, and DiagDO.
