This paper focuses on defending against a novel attack targeting PoW-based blockchains—namely, the Selfish Denial-of-Service (SDoS) attack. SDoS requires as little as 19.6% of the total computational power to earn revenue exceeding its proportional share. Alarmingly, as the revenue-to-cost ratio of Bitcoin mining has recently approached 1, SDoS attacks can easily induce all rational honest miners to voluntarily cease mining altogether. To defend SDoS, this paper proposes a game-based strategy to defend against the SDoS attack. This method restricts the strategy whereby a node publishes only the block header after finding a new block and later reveals the full private chain immediately after a competing block is observed. Experimental results show that the proposed method not only reduces the attacker’s revenue, thereby encouraging voluntary termination of the attack, but also mitigates the impact of the SDoS attack and increases the revenue of honest miners.
SNOVA is a post-quantum digital signature scheme based on multivariate quadratic equations. It can effectively resist quantum attacks. However, SNOVA faces computational efficiency bottlenecks in practical deployment due to the extensive matrix operations required during key generation. To address this issue, this paper proposes a GPU-accelerated scheme for SNOVA key generation. First, we adopt a multi-dimensional grid mapping strategy to optimize thread allocation. We design specialized 2D and 3D grid configurations to achieve efficient parallelization of matrix operations. Second, we design a three-tier memory optimization system that effectively reduces memory access latency. Finally, we construct a batch processing framework for key generation. Through pipeline and reuse mechanisms, we implement batch key generation for SNOVA. Experimental results demonstrate that compared to the basic CPU implementation, the key generation speed achieves up to 12.30 × improvement. In batch processing mode, GPU-SNOVA achieves up to 3.0 × improvement in key generation time compared to the AVX2 version, and up to 58.44 × improvement compared to the CPU version, meeting the requirements of high-throughput applications.
To address the issues of high communication overhead, unfair node election, and insufficient security in Byzantine Fault Tolerance (BFT)-based consensus algorithms in blockchain systems, this paper proposes a multidimensional reputation score and probabilistic election-based PBFT algorithm, termed MP-PBFT. A node reputation scoring mechanism is designed, evaluating historical behavior, participation level, and honesty, and employing a progressive reward-and-punishment strategy for dynamic scoring. This mechanism allows high-reputation nodes to be adaptively and dynamically selected for key roles, improving consensus efficiency while maintaining security. Furthermore, to overcome centralization tendencies of highest-score selection, a probabilistic election mechanism is introduced for critical nodes. By randomly selecting the advantages of high-reputation nodes and the participation opportunities of low-reputation nodes, honest behavior is encouraged and fairness is improved. Simulation experiments validate MP-PBFT’s performance in communication efficiency, election fairness, and security. Compared to traditional PBFT, MP-PBFT reduces communication overhead by 65
The growing demand for synthetic data in privacy-sensitive domains necessitates scheduling frameworks that balance generation efficiency with verifiable traceability. Existing approaches often struggle to maintain secure auditability while scaling across distributed environments. This paper presents TrustSched, a distributed scheduling framework integrating blockchain-backed integrity proofs for trusted synthetic data generation. By employing a Master-Executor architecture with slot-epoch scheduling, the system achieves parallelized data synthesis while enforcing hierarchical Merkle tree commitments for tamper-evident traceability. A batched blockchain submission mechanism decouples evidence anchoring from task execution, optimizing both computational throughput and on-chain efficiency. Experimental validation confirms the framework’s ability to maintain robust security guarantees without compromising synthesis performance, demonstrating its suitability for enterprise applications requiring auditable data generation.
Named Entity Recognition (NER) in classical Chinese is challenging due to ambiguous word boundaries, variant-standard character coexistence, and contextual polysemy. To address these issues, we propose a NER method based on the GuwenBERT-BiGRU-FOCAL-CRF framework enhanced with PGD adversarial training. GuwenBERT, pre-trained on classical Chinese corpora, provides character-level contextual representations, improving the capture of linguistic features. A Bidirectional Gated Recurrent Unit (BiGRU) models long-range dependencies, while a Conditional Random Field (CRF) captures label relations. Focal Loss optimizes training by emphasizing hard-to-classify samples. Robustness is further improved through adversarial training, where learnable perturbations are added to embeddings or intermediate features, enabling stable predictions under minor perturbations. Experiments show that our model outperforms the BERT-BiLSTM-CRF baseline on classical Chinese NER, with F1-score gains of 12.40
