Shiyin Lin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Shiyin Lin has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Bowen Jiang
University of the Chinese Academy of Sciences, Beijing, China
1 shared publication - Chenhui Qiang
Tencent, Beijing, China
1 shared publication - Jiekai Wu
Faculty of Health Data Science, Juntendo University, Tokyo, Japan
1 shared publication - Sen Nie
Chinese Academy of Sciences, Beijing, China
1 shared publication - Wenchao Ding
Tencent, Beijing, China
1 shared publication - Yanchao Hao
Tencent, Beijing, China
1 shared publication
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 11365-11373
DraftMind: Efficient Video Understanding in Vision-Language Models via Chain of Draft
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 11365-11373
Open AccessDraftMind: Efficient Video Understanding in Vision-Language Models via Chain of Draft
- Open Access
- 21 November 2025
- Published by MDPI AG in Preprints.org
Withdrawn: Optimizing Safety Alignment and Jailbreak Defense for Large Language Models
This preprint has been withdrawn at the request of the authors because institutional authorization was not obtained.
- 21 November 2025
- Published by MDPI AG in Preprints.org
Open AccessWithdrawn: Optimizing Safety Alignment and Jailbreak Defense for Large Language Models
This preprint has been withdrawn at the request of the authors because institutional authorization was not obtained.
- Open Access
- 14 October 2025
- Published by MDPI AG in Preprints.org
Research on the Vulnerability Identification Efficiency of Enhanced Reverse-Analyzed LLM Model in Binary Program Fuzz Testing
In response to the problems of low coverage, high blindness and missed detection of complex vulnerabilities in traditional fuzz testing for binary programs, this paper proposes an enhanced fuzz testing framework based on reverse analysis and large language model (LLM). This framework extracts control flow graphs and key path features through static reverse analysis, and captures input dependency trajectories through dynamic reverse analysis. These are then transformed into structured text input for LLM; the LLM, fine-tuned with binary vulnerability data, predicts high-risk areas and guides the fuzz testing module to generate targeted test cases. Experiments on the LAVA-M benchmark set and 10 actual closed-source binary programs show that this framework improves the vulnerability discovery rate by 34.2% compared to AFL, and shortens the average vulnerability discovery time by 28.7%. Ablation experiments verify the improvement effect of reverse analysis features on the semantic understanding of LLM (F1 value increases by 19.5%). The research confirms that the collaboration of reverse analysis and LLM can effectively enhance the vulnerability identification efficiency of binary program fuzz testing.
- 14 October 2025
- Published by MDPI AG in Preprints.org
Open AccessResearch on the Vulnerability Identification Efficiency of Enhanced Reverse-Analyzed LLM Model in Binary Program Fuzz Testing
In response to the problems of low coverage, high blindness and missed detection of complex vulnerabilities in traditional fuzz testing for binary programs, this paper proposes an enhanced fuzz testing framework based on reverse analysis and large language model (LLM). This framework extracts control flow graphs and key path features through static reverse analysis, and captures input dependency trajectories through dynamic reverse analysis. These are then transformed into structured text input for LLM; the LLM, fine-tuned with binary vulnerability data, predicts high-risk areas and guides the fuzz testing module to generate targeted test cases. Experiments on the LAVA-M benchmark set and 10 actual closed-source binary programs show that this framework improves the vulnerability discovery rate by 34.2% compared to AFL, and shortens the average vulnerability discovery time by 28.7%. Ablation experiments verify the improvement effect of reverse analysis features on the semantic understanding of LLM (F1 value increases by 19.5%). The research confirms that the collaboration of reverse analysis and LLM can effectively enhance the vulnerability identification efficiency of binary program fuzz testing.
- Open Access
- 23 September 2025
- Published by MDPI AG in Preprints.org
Hybrid Fuzzing with LLM-Guided Input Mutation and Semantic Feedback
Software fuzzing has become a cornerstone in automated vulnerability discovery, yet existing mutation strategies often lack semantic awareness, leading to redundant test cases and slow exploration of deep program states. In this work, we present a hybrid fuzzing framework that integrates static and dynamic analysis with Large Language Model (LLM)-guided input mutation and semantic feedback. Static analysis extracts control-flow and data-flow information, which is transformed into structured prompts for the LLM to generate syntactically valid and semantically diverse inputs. During execution, we augment traditional coverage-based feedback with semantic feedback signals—derived from program state changes, exception types, and output semantics—allowing the fuzzer to prioritize inputs that trigger novel program behaviors beyond mere code coverage. We implement our approach atop AFL++, combining program instrumentation with embedding-based semantic similarity metrics to guide seed selection. Evaluation on real-world open-source targets, including libpng, tcpdump, and sqlite, demonstrates that our method achieves faster time-to-first-bug, higher semantic diversity, and a competitive number of unique bugs compared to state-of-the-art fuzzers. This work highlights the potential of combining LLM reasoning with semantic-aware feedback to accelerate and deepen vulnerability discovery.
- 23 September 2025
- Published by MDPI AG in Preprints.org
Open AccessHybrid Fuzzing with LLM-Guided Input Mutation and Semantic Feedback
Software fuzzing has become a cornerstone in automated vulnerability discovery, yet existing mutation strategies often lack semantic awareness, leading to redundant test cases and slow exploration of deep program states. In this work, we present a hybrid fuzzing framework that integrates static and dynamic analysis with Large Language Model (LLM)-guided input mutation and semantic feedback. Static analysis extracts control-flow and data-flow information, which is transformed into structured prompts for the LLM to generate syntactically valid and semantically diverse inputs. During execution, we augment traditional coverage-based feedback with semantic feedback signals—derived from program state changes, exception types, and output semantics—allowing the fuzzer to prioritize inputs that trigger novel program behaviors beyond mere code coverage. We implement our approach atop AFL++, combining program instrumentation with embedding-based semantic similarity metrics to guide seed selection. Evaluation on real-world open-source targets, including libpng, tcpdump, and sqlite, demonstrates that our method achieves faster time-to-first-bug, higher semantic diversity, and a competitive number of unique bugs compared to state-of-the-art fuzzers. This work highlights the potential of combining LLM reasoning with semantic-aware feedback to accelerate and deepen vulnerability discovery.
- Open Access
- 11 September 2025
- Published by MDPI AG in Preprints.org
Abductive Inference in Retrieval‑Augmented Language Models: Generating and Validating Missing Premises
Large Language Models (LLMs) enhanced with retrieval---commonly referred to as Retrieval-Augmented Generation (RAG)---have demonstrated strong performance in knowledge-intensive tasks. However, RAG pipelines often fail when retrieved evidence is incomplete, leaving gaps in the reasoning process. In such cases, the process of generating plausible missing premises to explain observations---offers a principled approach to bridge these gaps. In this paper, we propose a framework that integrates abductive inference into retrieval-augmented LLMs. Our method detects insufficient evidence, generates candidate missing premises, and validates them through consistency and plausibility checks. Experimental results on abductive reasoning and multi-hop QA benchmarks show that our approach improves both answer accuracy and reasoning faithfulness. This work highlights abductive inference as a promising direction for enhancing the robustness and explainability of RAG systems.
- 11 September 2025
- Published by MDPI AG in Preprints.org
Open AccessAbductive Inference in Retrieval‑Augmented Language Models: Generating and Validating Missing Premises
Large Language Models (LLMs) enhanced with retrieval---commonly referred to as Retrieval-Augmented Generation (RAG)---have demonstrated strong performance in knowledge-intensive tasks. However, RAG pipelines often fail when retrieved evidence is incomplete, leaving gaps in the reasoning process. In such cases, the process of generating plausible missing premises to explain observations---offers a principled approach to bridge these gaps. In this paper, we propose a framework that integrates abductive inference into retrieval-augmented LLMs. Our method detects insufficient evidence, generates candidate missing premises, and validates them through consistency and plausibility checks. Experimental results on abductive reasoning and multi-hop QA benchmarks show that our approach improves both answer accuracy and reasoning faithfulness. This work highlights abductive inference as a promising direction for enhancing the robustness and explainability of RAG systems.
- Open Access
- 10 September 2025
- Published by MDPI AG in Preprints.org
LLM-Driven Adaptive Source–Sink Identification and False Positive Mitigation for Static Analysis
Static analysis is effective for discovering software vulnerabilities but notoriously suffers from incomplete source– sink specifications and excessive false positives (FPs). We present ADATAINT, an LLM-driven taint analysis framework that adap- tively infers source/sink specifications and filters spurious alerts through neuro-symbolic reasoning. Unlike LLM-only detectors, ADATAINT grounds model suggestions in program facts and con- straint validation, ensuring both adaptability and determinism. We evaluate ADATAINT on Juliet 1.3, SV-COMP-style C benchmarks, and three large real-world projects. Results show that ADATAINT reduces false positives by 43.7% on average and improves recall by 11.2% compared to state-of-the-art baselines (CodeQL, Joern, and LLM-only pipelines), while maintaining competitive runtime overhead. These findings demonstrate that combining LLM inference with symbolic validation offers a prac- tical path toward more accurate and reliable static vulnerability analysis.
- 10 September 2025
- Published by MDPI AG in Preprints.org
Open AccessLLM-Driven Adaptive Source–Sink Identification and False Positive Mitigation for Static Analysis
Static analysis is effective for discovering software vulnerabilities but notoriously suffers from incomplete source– sink specifications and excessive false positives (FPs). We present ADATAINT, an LLM-driven taint analysis framework that adap- tively infers source/sink specifications and filters spurious alerts through neuro-symbolic reasoning. Unlike LLM-only detectors, ADATAINT grounds model suggestions in program facts and con- straint validation, ensuring both adaptability and determinism. We evaluate ADATAINT on Juliet 1.3, SV-COMP-style C benchmarks, and three large real-world projects. Results show that ADATAINT reduces false positives by 43.7% on average and improves recall by 11.2% compared to state-of-the-art baselines (CodeQL, Joern, and LLM-only pipelines), while maintaining competitive runtime overhead. These findings demonstrate that combining LLM inference with symbolic validation offers a prac- tical path toward more accurate and reliable static vulnerability analysis.