DZDerui Zhu

Derui Zhu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Derui Zhu has a total of 49 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    DepWareTrans: Dependency-Aware Incremental Repository Migration across Co-executable Languages

    by Sivajeet Chand, Alexander Pretschner, Steve Haupt, Derui Zhu, Sushant Kumar Pandey

  • Open Access

    Stop When It Matters: Detectability-Guided Microbenchmarking for Performance Regression Testing

    by Zongxiong Chen, Derui Zhu, Jinfu Chen, Kundi Yao, Alexander Pretschner, Weiyi Shang, Manfred Hauswirth, Sonja Schimmler

  • Confidentiality and practicality of trusted execution environments for cross-organizational multi-model consistency checking

    by Sebastian Bergemann, Derui Zhu, Andreas Bayha, Mohammad Sadeghi, Colin Atkinson, Alexander Pretschner

  • Open Access

    LLM4JMH: Studying the Use of LLMs for Generating Java Performance Microbenchmarks

    by Zongxiong Chen, Derui Zhu, Kundi Yao, Weiyi Shang, Jinfu Chen, Jiahui Geng, Alexander Pretschner, Jens Grossklags, Manfred Hauswirth, Sonja Schimmler

  • Open Access

    Data Protection Vulnerabilities Assessment for Cross-Organizational Consistency Checking in TEE

    by Derui Zhu, Sebastian Bergemann, Mohammad Sadeghi, Colin Atkinson, Alexander Pretschner


  • Open Access

    Internal Activation Revision: Safeguarding Vision Language Models Without Parameter Update

    by Qing Li, Jiahui Geng, Derui Zhu, Zongxiong Chen, Kun Song, Lei Ma, Fakhri Karray

    Warning: This paper contains offensive content that may disturb some readers. Vision-language models (VLMs) demonstrate strong multimodal capabilities but have been found to be more susceptible to generating harmful content compared to their backbone large language models (LLMs). Our investigation reveals that the integration of images significantly shifts the model's internal activations during the forward pass, diverging from those triggered by textual input. Moreover, the safety alignments of LLMs embedded within VLMs are not sufficiently robust to handle the activations discrepancies, making the models vulnerable to even the simplest jailbreaking attacks. To address this issue, we propose an internal activation revision approach that efficiently revises activations during generation, steering the model toward safer outputs. Our framework incorporates revisions at both the layer and head levels, offering control over the model's generation at varying levels of granularity. In addition, we explore three strategies for constructing positive and negative samples and two approaches for extracting revision vectors, resulting in different variants of our method. Comprehensive experiments demonstrate that the internal activation revision method significantly improves the safety of widely used VLMs, reducing attack success rates by an average of 48.94%, 34.34%, 43.92%, and 52.98% on SafeBench, Safe-Unsafe, Unsafe, and MM-SafetyBench, respectively, while minimally impacting model helpfulness.


  • Open Access

    Safe Reinforcement Learning via Episodic Control

    by Zhuo Li, Derui Zhu, Jens Grossklags

    Safe reinforcement learning (Safe RL) aims to learn policies capable of learning and adapting within complex environments while ensuring actions remain free from catastrophic consequences. This is a critical consideration in domains such as robotics, autonomous vehicles, and healthcare. Unlike traditional RL, which focuses mainly on maximizing episodic rewards, Safe RL integrates safety constraints to balance rewards and safety. Current Safe RL algorithms, while promising, lack sample efficiency as they require extensive environmental interactions to perform multi-objective optimization. This paper introduces an episodic-control-based method to enhance sample efficiency in safe policy optimization. In RL, methods based on episodic control involve the storing and replaying of past experiences or episodes, aiding in more efficient and precise policy optimization. Our proposed method involves clustering, measuring, and storing previous states based on a joint metric of returns and safety. Subsequently, we retrieve these state measurements and incorporate them into the policy optimization process through reward shaping. This approach effectively guides the policy towards high-return and safe decisions. We evaluate the performance of our method on established Safe RL benchmarks, including six safety-critical agent control tasks. The results demonstrate that our method can concurrently achieve higher episodic returns and fewer violations of safety constraints compared to the baseline methods, suggesting an effective balance between earning rewards and safety.


  • Exploiting Internal Randomness for Privacy in Vertical Federated Learning

    by Yulian Sun, Li Duan, Ricardo Mendes, Derui Zhu, Yue Xia, Yong Li, Asja Fischer


  • LUNA: A Model-Based Universal Analysis Framework for Large Language Models

    by Da Song, Xuan Xie, Jiayang Song, Derui Zhu, Yuheng Huang, Felix Juefei-Xu, Lei Ma

    Over the past decade, Artificial Intelligence (AI) has had great success recently and is being used in a wide range of academic and industrial fields. More recently, Large Language Models (LLMs) have made rapid advancements that have propelled AI to a new level, enabling and empowering even more diverse applications and industrial domains with intelligence, particularly in areas like software engineering and natural language processing. Nevertheless, a number of emerging trustworthiness concerns and issues exhibited in LLMs, e.g., robustness and hallucination, have already recently received much attention, without properly solving which the widespread adoption of LLMs could be greatly hindered in practice. The distinctive characteristics of LLMs, such as the self-attention mechanism, extremely large neural network scale, and autoregressive generation usage contexts, differ from classic AI software based on Convolutional Neural Networks and Recurrent Neural Networks and present new challenges for quality analysis. Up to the present, it still lacks universal and systematic analysis techniques for LLMs despite the urgent industrial demand across diverse domains. Towards bridging such a gap, in this paper, we initiate an early exploratory study and propose a universal analysis framework for LLMs, named LUNA , which is designed to be general and extensible and enables versatile analysis of LLMs from multiple quality perspectives in a human-interpretable manner. In particular, we first leverage the data from desired trustworthiness perspectives to construct an abstract model as an auxiliary analysis asset and proxy, which is empowered by various abstract model construction methods built-in LUNA . To assess the quality of the abstract model, we collect and define a number of evaluation metrics, aiming at both the abstract model level and the semantics level. Then, the semantics, which is the degree of satisfaction of the LLM w.r.t. the trustworthiness perspective, is bound to and enriches the abstract model with semantics, which enables more detailed analysis applications for diverse purposes, e.g., abnormal behavior detection. To better understand the potential usefulness of our analysis framework LUNA, we conduct a large-scale evaluation, the results of which demonstrate that 1) the abstract model has the potential to distinguish normal and abnormal behavior in LLM, 2) LUNA is effective for the real-world analysis of LLMs in practice, and the hyperparameter settings influence the performance, 3) different evaluation metrics are in different correlations with the analysis performance. In order to encourage further studies in the quality assurance of LLMs, we made all of the code and more detailed experimental results data available on the supplementary website of this paper https://sites.google.com/view/llm-luna.


  • Open Access

    PoLLMgraph: Unraveling Hallucinations in Large Language Models via State Transition Dynamics

    by Derui Zhu, Dingfan Chen, Qing Li, Zongxiong Chen, Lei Ma, Jens Grossklags, Mario Fritz

    Despite tremendous advancements in large language models (LLMs) over recent years, a notably urgent challenge for their practical deployment is the phenomenon of hallucination, where the model fabricates facts and produces non-factual statements. In response, we propose PoLLMgraph, a Polygraph for LLMs, as an effective model-based white-box detection and forecasting approach. PoLLMgraph distinctly differs from the large body of existing research that concentrates on addressing such challenges through black-box evaluations. In particular, we demonstrate that hallucination can be effectively detected by analyzing the LLM's internal state transition dynamics during generation via tractable probabilistic models. Experimental results on various open-source LLMs confirm the efficacy of PoLLMgraph, outperforming state-of-the-art methods by a considerable margin, evidenced by over 20% improvement in AUC-ROC on common benchmarking datasets like TruthfulQA. Our work paves a new way for model-based white-box analysis of LLMs, motivating the research community to further explore, understand, and refine the intricate dynamics of LLM behaviors.