Selected work

Representative Papers

Examining the effectiveness of transformer-based smart contract vulnerability scan

Jan 01, 2026Journal of Systems and Software

This work addresses the critical need for accurate and efficient detection of security vulnerabilities in Ethereum smart contracts, which can lead to severe financial losses. We propose VASCOT, the first Transformer-based framework for vulnerability detection at the EVM bytecode level. By serializing bytecode and incorporating a sliding window mechanism, VASCOT effectively handles long input sequences, overcoming the length limitations of conventional models. A high-quality dataset, annotated based on execution traces, was constructed to support this approach. Evaluated on 16,469 real-world contracts deployed in 2022, VASCOT significantly outperforms LSTM-based baselines in detection performance and demonstrates superior generalization capability, establishing a new paradigm for smart contract security analysis.

2 citationsRead paper

Approval Integrity and Recovery in LLM Answer Publication

Sep 14, 2026

研究通过精确内容绑定、授权新鲜度和检查点恢复方法解决LLM系统发布完整性问题，对比了不同模型的性能并分析了错误来源。

0 citationsRead paper

From Density to Biopsy Decisions and Malignancy Prediction: A Benchmark Study of Multimodal Large Language Models Against Radiologists in Digital and Contrast-Enhanced Mammography

Sep 13, 2026

研究对比了四种多模态大型语言模型与放射科医生在乳腺密度评估、BI-RADS评估、活检候选确定及恶性概率估计上的表现，发现放射科医生在分类任务上更优，但某些模型在连续恶性概率估计中接近人类水平。

0 citationsRead paper

AccretionLink: On-Device Auditing of Exposure-Control Attacks on Attribute Inference

Aug 13, 2026

This study addresses the challenge of auditing private attribute inference under exposure control attacks by proposing a confidentiality-integrity security game model and an on-device trusted auditing framework. By integrating partial identification modeling for selection probabilities with dependency-aware e-processes, and leveraging Tensor G5 graph computing alongside P-256 signature authentication, the framework achieves efficient verification. Experimental results validate the effectiveness of attack advantage analysis, while successful deployment on the Pixel 10 platform demonstrates fallback-free encoding and key authentication mechanisms. Collectively, this work provides both formal guarantees and practical engineering solutions for ensuring privacy security in on-device AI systems against sophisticated inference attacks.

0 citationsRead paper

Accepted Prefixes Are Not All You Need: A Negative Result on PEFT-Based Block-Diffusion Drafting

Jul 14, 2026

This work investigates whether block-wise draft generation within a parameter-efficient fine-tuning (PEFT) framework can accelerate autoregressive language model inference. To this end, the authors propose PEFT-BD, which, for the first time, employs LoRA-style adapters as a block-wise draft generator to produce multi-token prefixes in parallel using the same backbone model, while the original backbone—without adapters—serves as the verifier. This design avoids tokenizer mismatches and eliminates the need for loading auxiliary models. Despite achieving high-quality prefix predictions with minimal trainable parameters and generating accepted prefixes of nontrivial length, the approach fails to yield end-to-end speedup. The key limitation lies in the draft phase still requiring a full forward pass through the backbone, incurring computational costs comparable to those of the verifier, thereby revealing that merely increasing accepted prefix length is insufficient for practical acceleration.

0 citationsRead paper

Recent publications

Latest Papers

Approval Integrity and Recovery in LLM Answer Publication

Sep 14, 2026

研究通过精确内容绑定、授权新鲜度和检查点恢复方法解决LLM系统发布完整性问题，对比了不同模型的性能并分析了错误来源。

0 citationsRead paper

From Density to Biopsy Decisions and Malignancy Prediction: A Benchmark Study of Multimodal Large Language Models Against Radiologists in Digital and Contrast-Enhanced Mammography

Sep 13, 2026

研究对比了四种多模态大型语言模型与放射科医生在乳腺密度评估、BI-RADS评估、活检候选确定及恶性概率估计上的表现，发现放射科医生在分类任务上更优，但某些模型在连续恶性概率估计中接近人类水平。

0 citationsRead paper

AccretionLink: On-Device Auditing of Exposure-Control Attacks on Attribute Inference

Aug 13, 2026

This study addresses the challenge of auditing private attribute inference under exposure control attacks by proposing a confidentiality-integrity security game model and an on-device trusted auditing framework. By integrating partial identification modeling for selection probabilities with dependency-aware e-processes, and leveraging Tensor G5 graph computing alongside P-256 signature authentication, the framework achieves efficient verification. Experimental results validate the effectiveness of attack advantage analysis, while successful deployment on the Pixel 10 platform demonstrates fallback-free encoding and key authentication mechanisms. Collectively, this work provides both formal guarantees and practical engineering solutions for ensuring privacy security in on-device AI systems against sophisticated inference attacks.

0 citationsRead paper

Accepted Prefixes Are Not All You Need: A Negative Result on PEFT-Based Block-Diffusion Drafting

Jul 14, 2026

This work investigates whether block-wise draft generation within a parameter-efficient fine-tuning (PEFT) framework can accelerate autoregressive language model inference. To this end, the authors propose PEFT-BD, which, for the first time, employs LoRA-style adapters as a block-wise draft generator to produce multi-token prefixes in parallel using the same backbone model, while the original backbone—without adapters—serves as the verifier. This design avoids tokenizer mismatches and eliminates the need for loading auxiliary models. Despite achieving high-quality prefix predictions with minimal trainable parameters and generating accepted prefixes of nontrivial length, the approach fails to yield end-to-end speedup. The key limitation lies in the draft phase still requiring a full forward pass through the backbone, incurring computational costs comparable to those of the verifier, thereby revealing that merely increasing accepted prefix length is insufficient for practical acceleration.

0 citationsRead paper

Conductance-Repair Evidence Graphs for Prospective Security Retrieval

Jul 04, 2026

This work addresses the challenge of forward-looking decision-making in secure retrieval scenarios where evidence information is asynchronous, missing, delayed, or contaminated. To tackle this, the authors propose a timestamped evidence graph framework that incorporates a time-aware acceptability mask for retrieval and introduces an admittance-based repair mechanism coupled with a repair certificate system. Missing channels are recursively restored via graph-stream propagation, while singular value decomposition (SVD) and einsum operations enable consistent kernel implementation and structured sparsity control across multiple backends (NumPy, PyTorch, JAX, TensorFlow). Theoretically, the study establishes a logarithmic lower bound for identifying missing channels, proves the NP-hardness of minimal-harm repair, and derives a fixed-parameter tractable bound for certifiable search over suspicious channels. Experiments demonstrate significant improvements in recall@k—from 0.017 to 0.069 on public datasets and up to 0.099 in synthetic settings—while ensuring verifiable repair processes and transparent backend usage logging.

0 citationsRead paper