Institution profile

Swinburne University of Technology

Academic institutionaustralasia · au
Official website
Research library140linked papers
Opportunities0open roles
Selected work

Representative Papers

A Systematic Literature Review on Large Language Models for Automated Program Repair

May 02, 2024arXiv.org

Research on large language models (LLMs) for automated program repair (APR) remains fragmented and lacks a systematic, unified understanding. Method: We conduct a systematic literature review (SLR) covering 127 papers published between 2020 and 2024, establishing the first comprehensive conceptual framework for LLM-based APR. We categorize model utilization strategies into three types—fine-tuning, prompt engineering, and hybrid ensemble—and perform multidimensional thematic analysis across input representation, semantic/security-specific repair scenarios, and open-science practices. Contribution/Results: We identify core challenges including model robustness, evaluation bias, and real-world deployment adaptability. The study yields a reusable taxonomy, benchmark insights, and methodological guidelines—delivering the APR community’s first holistic landscape map to precisely identify research gaps and inform future innovation pathways.

39 citations1 influentialRead paper

Rethinking Membership Inference Attacks Against Transfer Learning

Jan 20, 2025IEEE Transactions on Information Forensics and Security

In transfer learning, teacher models’ training data are vulnerable to membership inference attacks (MIAs), yet existing work predominantly assumes attackers have black-box or white-box access to the teacher model—overlooking privacy threats when only the student model is accessible in a white-box setting. Method: This paper identifies that representational discrepancies between teacher and student hidden layers can be exploited for MIAs, and proposes the first MIA framework requiring only white-box access to the student model. It employs shadow models to calibrate student-layer representations, models teacher–student representation divergence, and infers teacher training membership via reverse inference—without requiring the teacher model itself. Contribution/Results: The method achieves high inference accuracy across four benchmark datasets and diverse transfer learning tasks. It significantly expands the threat surface of MIAs in transfer learning and provides novel empirical evidence and insights for safeguarding teacher model privacy.

13 citationsRead paper

MORTAR: Metamorphic Multi-turn Testing for LLM-based Dialogue Systems

Dec 20, 2024arXiv.org

To address the long-standing test oracle problem in multi-turn LLM-based dialogue systems, this paper proposes the first knowledge graph–driven metamorphic testing framework for multi-turn dialogues. Methodologically: (1) it constructs dialogue-level knowledge graphs to model semantic structures across turns; and (2) it designs dialogue-level perturbations and metamorphic relations to enable reference-free, LLM-free automated test case generation. The key contribution is the first systematic adaptation of metamorphic testing to multi-turn dialogue scenarios—circumventing evaluation bias introduced by LLM-based oracles. Experiments on multiple mainstream LLM dialogue systems demonstrate that our approach detects significantly more unique defects than existing methods; notably, its detection rate for severe defects is four times higher than that of the best-performing single-turn metamorphic testing baseline, while maintaining low computational cost and high reliability.

2 citationsRead paper

Towards Effective Identification of Attack Techniques in Cyber Threat Intelligence Reports using Large Language Models

May 06, 2025

This paper addresses the challenge of identifying MITRE ATT&CK techniques in Cyber Threat Intelligence (CTI) reports, tackling three key issues: class imbalance, model overfitting, and domain-specific semantic complexity. We propose a novel two-stage paradigm: “LLM-based summarization preprocessing + SciBERT retraining.” First, open-source large language models (e.g., Llama2) generate high-quality synthetic samples to reconstruct a balanced training dataset. Second, a domain-adapted SciBERT model is fine-tuned on this enhanced data. Experiments demonstrate that our approach achieves F1-scores consistently exceeding 0.90 for critical ATT&CK technique identification—significantly outperforming baseline methods. The framework enhances automation efficiency in web-based CTI systems and strengthens human-AI collaborative defense. To our knowledge, this is the first work integrating LLM-driven data augmentation with lightweight domain-specific BERT for ATT&CK technique identification.

1 citationsRead paper

Ethical Concerns of Generative AI and Mitigation Strategies: A Systematic Mapping Study

Jan 08, 2025arXiv.org

This study systematically examines the multidimensional ethical challenges and cross-domain governance dilemmas arising from real-world deployments of generative AI—particularly large language models (LLMs). Method: Through a systematic literature review (SLR) and thematic coding, we structurally map 39 empirical studies using an original five-dimensional ethical framework. Contribution/Results: Our analysis uncovers a fundamental tension between the dynamic evolution of ethical risks and the persistent lag in governance responses—a finding not previously documented. We demonstrate that existing mitigation strategies exhibit severe adaptive deficits in high-stakes domains such as healthcare and public administration, stemming from misalignment among technological development, ethical reasoning, and institutional evolution. To address this, we propose a tripartite co-evolutionary pathway integrating ethics, technology, and institutions, offering both theoretical grounding and an actionable framework for resilient governance of generative AI.

1 citationsRead paper
Recent publications

Latest Papers