Institution profile

Griffith University

Academic institutionaustralasia · au
Official website
Research library353linked papers
Opportunities0open roles
Selected work

Representative Papers

Prompt Injection attack against LLM-integrated Applications

Jun 08, 2023arXiv.org

Prompt injection attacks pose an increasingly severe security threat to large language model (LLM) integrated applications, yet existing black-box attack methods suffer from limited practical efficacy. Method: This paper proposes HouYi—the first real-world-oriented, three-stage black-box prompt injection framework comprising pre-prompt injection, context-aware segmentation, and malicious payload delivery. HouYi uniquely enables automated triggering of high-impact consequences—including arbitrary LLM misuse and application-level prompt stealing—via black-box fuzzing, context-aware prompt engineering, and web-injection-inspired modeling. Contribution/Results: Evaluated through real-world penetration testing across 36 mainstream LLM applications, HouYi uncovered 31 critical vulnerabilities, independently confirmed by ten vendors—including Notion—with impact on millions of users. The work significantly advances LLM security practice by bridging the gap between theoretical attack models and deployable, scalable exploitation techniques.

536 citations38 influentialRead paper

Budgeted Embedding Table For Recommender Systems

Oct 23, 2023Web Search and Data Mining

To address memory overflow and poor scalability caused by fixed-size embeddings in recommender systems, this paper proposes a budget-controllable, table-level embedding optimization method. Unlike existing instance-wise search approaches or manually tuned lightweight embedding schemes, our method introduces a novel table-level action modeling and set representation learning framework, incorporating an action suitability predictor to jointly optimize embedding dimensions across all feature tables—globally, efficiently, and strictly under user-specified memory constraints. By integrating set neural networks with memory-aware architecture search, the approach eliminates the need for heuristic trade-off coefficients. Evaluated on two real-world datasets with three mainstream recommendation models, our method achieves state-of-the-art performance, significantly improving both recommendation accuracy and memory efficiency while rigorously satisfying multiple predefined memory budgets.

10 citationsRead paper

Scalable Dynamic Embedding Size Search for Streaming Recommendation

Jul 22, 2024International Conference on Information and Knowledge Management

To address the unbounded embedding storage overhead caused by continuously growing users and items in streaming recommendation, this paper proposes Scalable Lightweight Embeddings (SCALL), which adaptively allocates embedding dimensions per entity under strict memory budget constraints. Its core contributions are threefold: (1) the first probabilistic distribution-based embedding size sampling mechanism, enabling precise memory control; (2) a reinforcement learning search paradigm with fixed-length state representations, supporting dynamic size generation for unseen entities; and (3) mean-pooling-based state modeling coupled with dynamic memory-aware optimization. Experiments on two public streaming recommendation datasets demonstrate that SCALL reduces storage overhead by up to 62% while improving recommendation accuracy—Recall@10 increases by 3.1–5.7%. To our knowledge, SCALL is the first method to jointly optimize memory controllability and model performance in streaming recommendation.

9 citations1 influentialRead paper

ANUBIS: Skeleton Action Recognition Dataset, Review, and Benchmark

May 04, 2022arXiv.org

Existing 3D skeleton-based action recognition research suffers from fragmented representation taxonomies and evaluation protocols misaligned with real-world scenarios; moreover, mainstream datasets lack critical dimensions—including rear-view perspectives, multi-person interactions, fine-grained or violent actions, and pandemic-era behaviors. To address these gaps, we propose a four-dimensional taxonomy (dataset design, spatial modeling, temporal modeling, and signal enhancement) and introduce ANUBIS: the first large-scale, multi-view 3D skeleton dataset explicitly designed for realistic challenges. ANUBIS features rear-view captures, 101 action classes (including 21 pandemic-related behaviors), and standardized recordings from 128 participants using Azure Kinect’s multi-sensor fusion. We further establish a unified benchmark framework, enabling reproducible evaluation of 12 state-of-the-art models. Our analysis identifies temporal modeling capacity and signal robustness as the primary bottlenecks limiting current performance.

4 citationsRead paper

Agent Skills in the Wild: An Empirical Study of Security Vulnerabilities at Scale

Jan 15, 2026

This study addresses the critical lack of systematic security auditing in the current ecosystem of AI agent skills, which harbors widespread yet underrecognized security risks. The authors present the first security taxonomy for agent skills grounded in real-world vulnerabilities and introduce SkillScan, a multi-stage detection framework that integrates static code analysis with large language model–based semantic classification. Empirical evaluation across 42,447 skills from two major marketplaces reveals that 26.1% contain vulnerabilities, with data leakage and privilege escalation being the most prevalent; executable-script skills exhibit significantly higher risk. The proposed method achieves 86.7% precision and 82.5% recall in vulnerability detection. The dataset and toolkit are publicly released to support further research.

3 citationsRead paper
Recent publications

Latest Papers