arXiv subjects

Chunming Wu

Publications and source records attributed to Chunming Wu.

At least 19 recordsLinked to original sources

arXiv2026

DreamGuard: Efficient Runtime Guardrail for LLM Agents via Risk-Aware World Model

As large language model (LLM) agents increasingly invoke external tools and interact with real-world systems, unsafe actions may cause irreversible consequences on external states, user data, and downstream services. Recent runtime guardrails mitigate such risks by checking proposed actions before execution, but many remain reactive: they primarily assess the apparent safety of the current action, lacking an explicit model of how risk evolves across the trajectory. This limitation creates a critical blind spot for long-horizon risks, where individually benign-looking actions can gradually drift the agent toward hazardous states. In response, we propose DreamGuard, a proactive guardrail for LLM agents built around a risk-aware world model. The world model maintains a compact recurrent latent state over the trajectory and predicts future latent states from which DreamGuard derives immediate-hazard and prefix-risk evidence. It then fuses these multi-horizon signals into intervention decisions before execution. Experiments across four benchmarks and an online guardrail evaluation show that DreamGuard outperforms generic, reactive, and proactive guardrail baselines, achieves the best safety-utility trade-off among evaluated guardrails, and maintains an average end-to-end latency of 25 ms per call.

cs.AI↗

arXiv2026

Evidence-in-the-Loop: Trace-Driven Optimization for Customer-Service LLM Agents

Production customer-service bots must improve answer quality across iterative releases, yet large language models must not bypass evidence boundaries, policy rules, or human-handoff safeguards. We present an \textbf{Evidence-Grounded Customer-Service Agent Workflow} deployed in a real-world customer-service setting. BM25 recall, issue-title-vector recall, issue-description-vector recall, weighted RRF fusion, and cross-encoder reranking construct grounded FAQ evidence for controlled LLM decisions. Policy-guided orchestration then combines this RAG evidence with scenario-specific rule evidence, conversation memory, and clarification state inside a fixed LangGraph DAG~\cite{langgraph2024}. The paper contributes three reusable deployment patterns: \textbf{hybrid RAG evidence construction}, where multi-channel retrieval and reranking produce auditable FAQ candidates; \textbf{evidence-grounded issue/action decision}, where an Evidence-Grounded Decision Module selects an issue/action from typed FAQ evidence and scenario-specific rule evidence; and \textbf{trace-driven RAG and reranker improvement}, where traces diagnose whether failures come from recall, ranking, final candidate selection, clarification, rule-derived evidence, or action policy, and where reranker fine-tuning is evaluated not only for in-domain gain but also for forgetting risk.

cs.IR↗

arXiv2026

OmniPlan: An Adaptive Framework for Timely and Near-Optimal Network Planning Optimization

Network planning optimization is a fundamental problem across diverse domains, including transportation systems, communication networks, and power grids. It requires simultaneous optimization of multiple competing objectives under complex constraints. Existing network planning optimization frameworks rely on mixed integer programming (MIP) solvers, heuristics, and deep reinforcement learning (DRL) models to compute planning decisions. However, they lack effective adaptability to diverse and dynamic user intents, thus leading to the trade-off between execution time and optimality. In this paper, we propose OmniPlan, an adaptive framework that achieves both timeliness and near-optimality in network planning optimization. To achieve the adaptability lacking in existing solutions, OmniPlan employs a large language model (LLM)-based interpreter to convert heterogeneous natural-language intents into a unified and quantifiable user-preference vector. Then it employs a mixture-of-experts architecture that integrates MIP solvers, heuristics, and DRL models as specialized experts, where OmniPlan adapts to diverse intents by dynamically selecting timely and near-optimal experts. Finally, it incorporates a DRL-based expert configuration module that fine-tunes optimization objective weights to align planning decisions with user-specific preferences. We evaluate OmniPlan with a representative real-world workload, i.e., distributed machine learning (ML), where we leverage OmniPlan to offload a wide spectrum of ML inference tasks, e.g., decision trees, SVM, naive Bayes, XGBoost, and random forests, onto a network of hardware devices. Our experiments on a real-world testbed indicate that OmniPlan achieves near-optimal and low-execution-time offloading for real-world ML inference tasks, reducing latency by up to 97.8\% and network device resource consumption by up to 11.5\%.

cs.NI↗

arXiv2026

ICON: Intent-Context Coupling for Efficient Multi-Turn Jailbreak Attack

Multi-turn jailbreak attacks have emerged as a critical threat to Large Language Models (LLMs), bypassing safety mechanisms by progressively constructing adversarial contexts from scratch and incrementally refining prompts. However, existing methods suffer from the inefficiency of incremental context construction that requires step-by-step LLM interaction, and often stagnate in suboptimal regions due to surface-level optimization. In this paper, we characterize the Intent-Context Coupling phenomenon, revealing that LLM safety constraints are significantly relaxed when a malicious intent is coupled with a semantically congruent context pattern. Driven by this insight, we propose ICON, an automated multi-turn jailbreak framework that efficiently constructs an authoritative-style context via prior-guided semantic routing. Specifically, ICON first routes the malicious intent to a congruent context pattern (e.g., Scientific Research) and instantiates it into an attack prompt sequence. This sequence progressively builds the authoritative-style context and ultimately elicits prohibited content. In addition, ICON incorporates a Hierarchical Optimization Strategy that combines local prompt refinement with global context switching, preventing the attack from stagnating in ineffective contexts. Experimental results across eight SOTA LLMs demonstrate the effectiveness of ICON, achieving a state-of-the-art average Attack Success Rate (ASR) of 97.1\%. Code is available at https://github.com/xwlin-roy/ICON.

cs.CR↗

arXiv2025

Breaking Isolation: A New Perspective on Hypervisor Exploitation via Cross-Domain Attacks

Hypervisors are under threat by critical memory safety vulnerabilities, with pointer corruption being one of the most prevalent and severe forms. Existing exploitation frameworks depend on identifying highly-constrained structures in the host machine and accurately determining their runtime addresses, which is ineffective in hypervisor environments where such structures are rare and further obfuscated by Address Space Layout Randomization (ASLR). We instead observe that modern virtualization environments exhibit weak memory isolation -- guest memory is fully attacker-controlled yet accessible from the host, providing a reliable primitive for exploitation. Based on this observation, we present the first systematic characterization and taxonomy of Cross-Domain Attacks (CDA), a class of exploitation techniques that enable capability escalation through guest memory reuse. To automate this process, we develop a system that identifies cross-domain gadgets, matches them with corrupted pointers, synthesizes triggering inputs, and assembles complete exploit chains. Our evaluation on 15 real-world vulnerabilities across QEMU and VirtualBox shows that CDA is widely applicable and effective.

cs.CR↗

arXiv2025

A Survey of LLM-Driven AI Agent Communication: Protocols, Security Risks, and Defense Countermeasures

In recent years, Large-Language-Model-driven AI agents have exhibited unprecedented intelligence and adaptability. Nowadays, agents are undergoing a new round of evolution. They no longer act as an isolated island like LLMs. Instead, they start to communicate with diverse external entities, such as other agents and tools, to perform complex tasks. Under this trend, agent communication is regarded as a foundational pillar of the next communication era, and many organizations have intensively begun to design related communication protocols (e.g., Anthropic's MCP and Google's A2A) within the past year. However, this new field exposes significant security hazards, which can cause severe damage to real-world scenarios. To help researchers quickly figure out this promising topic and benefit the future agent communication development, this paper presents a comprehensive survey of agent communication security. More precisely, we present the first clear definition of agent communication. Besides, we propose a framework that categorizes agent communication into three classes and uses a three-layered communication architecture to illustrate how each class works. Next, for each communication class, we dissect related communication protocols and analyze the security risks, illustrating which communication layer the risks arise from. Then, we provide an outlook on the possible defense countermeasures for each risk. In addition, we conduct experiments using MCP and A2A to help readers better understand the novel vulnerabilities brought by agent communication. Finally, we discuss open issues and future directions in this promising research field. We also publish a repository that maintains a list of related papers on https://github.com/theshi-1128/awesome-agent-communication-security.

cs.CR↗

arXiv2025

LargeMvC-Net: Anchor-based Deep Unfolding Network for Large-scale Multi-view Clustering

Deep anchor-based multi-view clustering methods enhance the scalability of neural networks by utilizing representative anchors to reduce the computational complexity of large-scale clustering. Despite their scalability advantages, existing approaches often incorporate anchor structures in a heuristic or task-agnostic manner, either through post-hoc graph construction or as auxiliary components for message passing. Such designs overlook the core structural demands of anchor-based clustering, neglecting key optimization principles. To bridge this gap, we revisit the underlying optimization problem of large-scale anchor-based multi-view clustering and unfold its iterative solution into a novel deep network architecture, termed LargeMvC-Net. The proposed model decomposes the anchor-based clustering process into three modules: RepresentModule, NoiseModule, and AnchorModule, corresponding to representation learning, noise suppression, and anchor indicator estimation. Each module is derived by unfolding a step of the original optimization procedure into a dedicated network component, providing structural clarity and optimization traceability. In addition, an unsupervised reconstruction loss aligns each view with the anchor-induced latent space, encouraging consistent clustering structures across views. Extensive experiments on several large-scale multi-view benchmarks show that LargeMvC-Net consistently outperforms state-of-the-art methods in terms of both effectiveness and scalability.

cs.CV↗

arXiv2024

G-Fuzz: A Directed Fuzzing Framework for gVisor

gVisor is a Google-published application-level kernel for containers. As gVisor is lightweight and has sound isolation, it has been widely used in many IT enterprises \cite{Stripe, DigitalOcean, Cloundflare}. When a new vulnerability of the upstream gVisor is found, it is important for the downstream developers to test the corresponding code to maintain the security. To achieve this aim, directed fuzzing is promising. Nevertheless, there are many challenges in applying existing directed fuzzing methods for gVisor. The core reason is that existing directed fuzzers are mainly for general C/C++ applications, while gVisor is an OS kernel written in the Go language. To address the above challenges, we propose G-Fuzz, a directed fuzzing framework for gVisor. There are three core methods in G-Fuzz, including lightweight and fine-grained distance calculation, target related syscall inference and utilization, and exploration and exploitation dynamic switch. Note that the methods of G-Fuzz are general and can be transferred to other OS kernels. We conduct extensive experiments to evaluate the performance of G-Fuzz. Compared to Syzkaller, the state-of-the-art kernel fuzzer, G-Fuzz outperforms it significantly. Furthermore, we have rigorously evaluated the importance for each core method of G-Fuzz. G-Fuzz has been deployed in industry and has detected multiple serious vulnerabilities.

cs.CR↗

arXiv2024

An Attention-Based Deep Generative Model for Anomaly Detection in Industrial Control Systems

Anomaly detection is critical for the secure and reliable operation of industrial control systems. As our reliance on such complex cyber-physical systems grows, it becomes paramount to have automated methods for detecting anomalies, preventing attacks, and responding intelligently. {This paper presents a novel deep generative model to meet this need. The proposed model follows a variational autoencoder architecture with a convolutional encoder and decoder to extract features from both spatial and temporal dimensions. Additionally, we incorporate an attention mechanism that directs focus towards specific regions, enhancing the representation of relevant features and improving anomaly detection accuracy. We also employ a dynamic threshold approach leveraging the reconstruction probability and make our source code publicly available to promote reproducibility and facilitate further research. Comprehensive experimental analysis is conducted on data from all six stages of the Secure Water Treatment (SWaT) testbed, and the experimental results demonstrate the superior performance of our approach compared to several state-of-the-art baseline techniques.

cs.LG↗

arXiv2024

AdvSQLi: Generating Adversarial SQL Injections against Real-world WAF-as-a-service

As the first defensive layer that attacks would hit, the web application firewall (WAF) plays an indispensable role in defending against malicious web attacks like SQL injection (SQLi). With the development of cloud computing, WAF-as-a-service, as one kind of Security-as-a-service, has been proposed to facilitate the deployment, configuration, and update of WAFs in the cloud. Despite its tremendous popularity, the security vulnerabilities of WAF-as-a-service are still largely unknown, which is highly concerning given its massive usage. In this paper, we propose a general and extendable attack framework, namely AdvSQLi, in which a minimal series of transformations are performed on the hierarchical tree representation of the original SQLi payload, such that the generated SQLi payloads can not only bypass WAF-as-a-service under black-box settings but also keep the same functionality and maliciousness as the original payload. With AdvSQLi, we make it feasible to inspect and understand the security vulnerabilities of WAFs automatically, helping vendors make products more secure. To evaluate the attack effectiveness and efficiency of AdvSQLi, we first employ two public datasets to generate adversarial SQLi payloads, leading to a maximum attack success rate of 100% against state-of-the-art ML-based SQLi detectors. Furthermore, to demonstrate the immediate security threats caused by AdvSQLi, we evaluate the attack effectiveness against 7 WAF-as-a-service solutions from mainstream vendors and find all of them are vulnerable to AdvSQLi. For instance, AdvSQLi achieves an attack success rate of over 79% against the F5 WAF. Through in-depth analysis of the evaluation results, we further condense out several general yet severe flaws of these vendors that cannot be easily patched.

cs.CR↗

arXiv2023

Packet-Level Adversarial Network Traffic Crafting using Sequence Generative Adversarial Networks

The surge in the internet of things (IoT) devices seriously threatens the current IoT security landscape, which requires a robust network intrusion detection system (NIDS). Despite superior detection accuracy, existing machine learning or deep learning based NIDS are vulnerable to adversarial examples. Recently, generative adversarial networks (GANs) have become a prevailing method in adversarial examples crafting. However, the nature of discrete network traffic at the packet level makes it hard for GAN to craft adversarial traffic as GAN is efficient in generating continuous data like image synthesis. Unlike previous methods that convert discrete network traffic into a grayscale image, this paper gains inspiration from SeqGAN in sequence generation with policy gradient. Based on the structure of SeqGAN, we propose Attack-GAN to generate adversarial network traffic at packet level that complies with domain constraints. Specifically, the adversarial packet generation is formulated into a sequential decision making process. In this case, each byte in a packet is regarded as a token in a sequence. The objective of the generator is to select a token to maximize its expected end reward. To bypass the detection of NIDS, the generated network traffic and benign traffic are classified by a black-box NIDS. The prediction results returned by the NIDS are fed into the discriminator to guide the update of the generator. We generate malicious adversarial traffic based on a real public available dataset with attack functionality unchanged. The experimental results validate that the generated adversarial samples are able to deceive many existing black-box NIDS.

cs.CR↗

arXiv2023

Self-Supervised Interest Transfer Network via Prototypical Contrastive Learning for Recommendation

Cross-domain recommendation has attracted increasing attention from industry and academia recently. However, most existing methods do not exploit the interest invariance between domains, which would yield sub-optimal solutions. In this paper, we propose a cross-domain recommendation method: Self-supervised Interest Transfer Network (SITN), which can effectively transfer invariant knowledge between domains via prototypical contrastive learning. Specifically, we perform two levels of cross-domain contrastive learning: 1) instance-to-instance contrastive learning, 2) instance-to-cluster contrastive learning. Not only that, we also take into account users' multi-granularity and multi-view interests. With this paradigm, SITN can explicitly learn the invariant knowledge of interest clusters between domains and accurately capture users' intents and preferences. We conducted extensive experiments on a public dataset and a large-scale industrial dataset collected from one of the world's leading e-commerce corporations. The experimental results indicate that SITN achieves significant improvements over state-of-the-art recommendation methods. Additionally, SITN has been deployed on a micro-video recommendation platform, and the online A/B testing results further demonstrate its practical value. Supplement is available at: https://github.com/fanqieCoffee/SITN-Supplement.

cs.IR↗

arXiv2023

Adversarial Attacks against Windows PE Malware Detection: A Survey of the State-of-the-Art

Malware has been one of the most damaging threats to computers that span across multiple operating systems and various file formats. To defend against ever-increasing and ever-evolving malware, tremendous efforts have been made to propose a variety of malware detection that attempt to effectively and efficiently detect malware so as to mitigate possible damages as early as possible. Recent studies have shown that, on the one hand, existing ML and DL techniques enable superior solutions in detecting newly emerging and previously unseen malware. However, on the other hand, ML and DL models are inherently vulnerable to adversarial attacks in the form of adversarial examples. In this paper, we focus on malware with the file format of portable executable (PE) in the family of Windows operating systems, namely Windows PE malware, as a representative case to study the adversarial attack methods in such adversarial settings. To be specific, we start by first outlining the general learning framework of Windows PE malware detection based on ML/DL and subsequently highlighting three unique challenges of performing adversarial attacks in the context of Windows PE malware. Then, we conduct a comprehensive and systematic review to categorize the state-of-the-art adversarial attacks against PE malware detection, as well as corresponding defenses to increase the robustness of Windows PE malware detection. Finally, we conclude the paper by first presenting other related attacks against Windows PE malware detection beyond the adversarial attacks and then shedding light on future research directions and opportunities. In addition, a curated resource list of adversarial attacks and defenses for Windows PE malware detection is also available at https://github.com/ryderling/adversarial-attacks-and-defenses-for-windows-pe-malware-detection.

cs.CR↗

arXiv2022

Towards the Desirable Decision Boundary by Moderate-Margin Adversarial Training

Adversarial training, as one of the most effective defense methods against adversarial attacks, tends to learn an inclusive decision boundary to increase the robustness of deep learning models. However, due to the large and unnecessary increase in the margin along adversarial directions, adversarial training causes heavy cross-over between natural examples and adversarial examples, which is not conducive to balancing the trade-off between robustness and natural accuracy. In this paper, we propose a novel adversarial training scheme to achieve a better trade-off between robustness and natural accuracy. It aims to learn a moderate-inclusive decision boundary, which means that the margins of natural examples under the decision boundary are moderate. We call this scheme Moderate-Margin Adversarial Training (MMAT), which generates finer-grained adversarial examples to mitigate the cross-over problem. We also take advantage of logits from a teacher model that has been well-trained to guide the learning of our model. Finally, MMAT achieves high natural accuracy and robustness under both black-box and white-box attacks. On SVHN, for example, state-of-the-art robustness and natural accuracy are achieved.

cs.CV↗

arXiv2022

Treating Crowdsourcing as Examination: How to Score Tasks and Online Workers?

Crowdsourcing is an online outsourcing mode which can solve the current machine learning algorithm's urge need for massive labeled data. Requester posts tasks on crowdsourcing platforms, which employ online workers over the Internet to complete tasks, then aggregate and return results to requester. How to model the interaction between different types of workers and tasks is a hot spot. In this paper, we try to model workers as four types based on their ability: expert, normal worker, sloppy worker and spammer, and divide tasks into hard, medium and easy task according to their difficulty. We believe that even experts struggle with difficult tasks while sloppy workers can get easy tasks right, and spammers always give out wrong answers deliberately. So, good examination tasks should have moderate degree of difficulty and discriminability to score workers more objectively. Thus, we first score workers' ability mainly on the medium difficult tasks, then reducing the weight of answers from sloppy workers and modifying the answers from spammers when inferring the tasks' ground truth. A probability graph model is adopted to simulate the task execution process, and an iterative method is adopted to calculate and update the ground truth, the ability of workers and the difficulty of the task successively. We verify the rightness and effectiveness of our algorithm both in simulated and real crowdsourcing scenes.

cs.HC↗

arXiv2021

Multilevel Graph Matching Networks for Deep Graph Similarity Learning

While the celebrated graph neural networks yield effective representations for individual nodes of a graph, there has been relatively less success in extending to the task of graph similarity learning. Recent work on graph similarity learning has considered either global-level graph-graph interactions or low-level node-node interactions, however ignoring the rich cross-level interactions (e.g., between each node of one graph and the other whole graph). In this paper, we propose a multi-level graph matching network (MGMN) framework for computing the graph similarity between any pair of graph-structured objects in an end-to-end fashion. In particular, the proposed MGMN consists of a node-graph matching network for effectively learning cross-level interactions between each node of one graph and the other whole graph, and a siamese graph neural network to learn global-level interactions between two input graphs. Furthermore, to compensate for the lack of standard benchmark datasets, we have created and collected a set of datasets for both the graph-graph classification and graph-graph regression tasks with different sizes in order to evaluate the effectiveness and robustness of our models. Comprehensive experiments demonstrate that MGMN consistently outperforms state-of-the-art baseline models on both the graph-graph classification and graph-graph regression tasks. Compared with previous work, MGMN also exhibits stronger robustness as the sizes of the two input graphs increase.

cs.LG↗

arXiv2021

STEP: Spatial-Temporal Network Security Event Prediction

Network security events prediction helps network operators to take response strategies from a proactive perspective, and reduce the cost caused by network attacks, which is of great significance for maintaining the security of the entire network. Most of the existing event prediction methods rely on temporal characteristics and are dedicated to exploring time series predictions, but ignoring the spatial relationship between hosts. This paper combines the temporal and spatial characteristics of security events and proposes a spatial-temporal event prediction model, named STEP. In particular, STEP formulates the security events prediction into a spatial-temporal sequence prediction. STEP utilizes graph convolution operation to capture the spatial characteristics of hosts in the network, and adopts the long short term memory (LSTM) to capture the dynamic temporal dependency of events. This paper verifies the proposed STEP scheme on two public data sets. The experimental results show that the prediction accuracy of security events under STEP is higher than that of benchmark models such as LSTM, ConvLSTM. Besides, STEP achieves high prediction accuracy when we predict events from different lengths of sequence.

cs.CR↗

arXiv2021

Deep Graph Matching and Searching for Semantic Code Retrieval

Code retrieval is to find the code snippet from a large corpus of source code repositories that highly matches the query of natural language description. Recent work mainly uses natural language processing techniques to process both query texts (i.e., human natural language) and code snippets (i.e., machine programming language), however neglecting the deep structured features of query texts and source codes, both of which contain rich semantic information. In this paper, we propose an end-to-end deep graph matching and searching (DGMS) model based on graph neural networks for the task of semantic code retrieval. To this end, we first represent both natural language query texts and programming language code snippets with the unified graph-structured data, and then use the proposed graph matching and searching model to retrieve the best matching code snippet. In particular, DGMS not only captures more structural information for individual query texts or code snippets but also learns the fine-grained similarity between them by cross-attention based semantic matching operations. We evaluate the proposed DGMS model on two public code retrieval datasets with two representative programming languages (i.e., Java and Python). Experiment results demonstrate that DGMS significantly outperforms state-of-the-art baseline models by a large margin on both datasets. Moreover, our extensive ablation studies systematically investigate and illustrate the impact of each part of DGMS.

cs.AI↗