Communication topology is a critical factor in the utility and safety of LLM-based multi-agent systems (LLM-MAS), making it a high-value intellectual property (IP) whose confidentiality remains insufficiently studied. % Existing topology inference attempts rely on impractical assumptions, including control over the administrative agent and direct identity queries via jailbreaks, which are easily defeated by basic keyword-based defenses. As a result, prior analyses fail to capture the real-world threat of such attacks. % To bridge this realism gap, we propose \textit{WebWeaver}, an attack framework that infers the complete LLM-MAS topology by compromising only a single arbitrary agent instead of the administrative agent. % Unlike prior approaches, WebWeaver relies solely on agent contexts rather than agent IDs, enabling significantly stealthier inference. % WebWeaver further introduces a new covert jailbreak-based mechanism and a novel fully jailbreak-free diffusion design to handle cases where jailbreaks fail. % Additionally, we address a key challenge in diffusion-based inference by proposing a masking strategy that preserves known topology during diffusion, with theoretical guarantees of correctness. % Extensive experiments show that WebWeaver substantially outperforms state-of-the-art (SOTA) baselines, achieving about 60\% higher inference accuracy under active defenses with negligible overhead.

Pruning-based unlearning has recently emerged as a fast, training-free, and data-independent approach to remove undesired concepts from diffusion models. It promises high efficiency and robustness, offering an attractive alternative to traditional fine-tuning or editing-based unlearning. However, in this paper we uncover a hidden danger behind this promising paradigm. We find that the locations of pruned weights, typically set to zero during unlearning, can act as side-channel signals that leak critical information about the erased concepts. To verify this vulnerability, we design a novel attack framework capable of reviving erased concepts from pruned diffusion models in a fully data-free and training-free manner. Our experiments confirm that pruning-based unlearning is not inherently secure, as erased concepts can be effectively revived without any additional data or retraining. Extensive experiments on diffusion-based unlearning based on concept related weights lead to the conclusion: once the critical concept-related weights in diffusion models are identified, our method can effectively recover the original concept regardless of how the weights are manipulated. Finally, we explore potential defense strategies and advocate safer pruning mechanisms that conceal pruning locations while preserving unlearning effectiveness, providing practical insights for designing more secure pruning-based unlearning frameworks.

Current quantum neural networks suffer from extreme sensitivity to both adversarial perturbations and hardware noise, creating a significant barrier to real-world deployment. Existing robustness techniques typically sacrifice clean accuracy or require prohibitive computational resources. We propose a hybrid quantum-classical Differentiable Quantum Architecture Search (DQAS) framework that addresses these limitations by jointly optimizing circuit structure and robustness through gradient-based methods. Our approach enhances traditional DQAS with a lightweight Classical Noise Layer applied before quantum processing, enabling simultaneous optimization of gate selection and noise parameters. This design preserves the quantum circuit's integrity while introducing trainable perturbations that enhance robustness without compromising standard performance. Experimental validation on MNIST, FashionMNIST, and CIFAR datasets shows consistent improvements in both clean and adversarial accuracy compared to existing quantum architecture search methods. Under various attack scenarios, including Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), Basic Iterative Method (BIM), and Momentum Iterative Method (MIM), and under realistic quantum noise conditions, our hybrid framework maintains superior performance. Testing on actual quantum hardware confirms the practical viability of discovered architectures. These results demonstrate that strategic classical preprocessing combined with differentiable quantum architecture optimization can significantly enhance quantum neural network robustness while maintaining computational efficiency.

The increasing deployment of large language models (LLMs) in safety-critical applications raises fundamental challenges in systematically evaluating robustness against adversarial behaviors. Existing red-teaming practices are largely manual and expert-driven, which limits scalability, reproducibility, and coverage in high-dimensional prompt spaces. We formulate automated LLM red-teaming as a structured adversarial search problem and propose a learning-driven framework for scalable vulnerability discovery. The approach combines meta-prompt-guided adversarial prompt generation with a hierarchical execution and detection pipeline, enabling standardized evaluation across six representative threat categories, including reward hacking, deceptive alignment, data exfiltration, sandbagging, inappropriate tool use, and chain-of-thought manipulation. Extensive experiments on GPT-OSS-20B identify 47 vulnerabilities, including 21 high-severity failures and 12 previously undocumented attack patterns. Compared with manual red-teaming under matched query budgets, our method achieves a 3.9$\times$ higher discovery rate with 89\% detection accuracy, demonstrating superior coverage, efficiency, and reproducibility for large-scale robustness evaluation.

Given the high cost of large language model (LLM) training from scratch, safeguarding LLM intellectual property (IP) has become increasingly crucial. As the standard paradigm for IP ownership verification, LLM fingerprinting thus plays a vital role in addressing this challenge. Existing LLM fingerprinting methods verify ownership by extracting or injecting model-specific features. However, they overlook potential attacks during the verification process, leaving them ineffective when the model thief fully controls the LLM's inference process. In such settings, attackers may share prompt-response pairs to enable fingerprint unlearning or manipulate outputs to evade exact-match verification. We propose iSeal, the first fingerprinting method designed for reliable verification when the model thief controls the suspected LLM in an end-to-end manner. It injects unique features into both the model and an external module, reinforced by an error-correction mechanism and a similarity-based verification strategy. These components are resistant to verification-time attacks, including collusion-based fingerprint unlearning and response manipulation, backed by both theoretical analysis and empirical results. iSeal achieves 100 percent Fingerprint Success Rate (FSR) on 12 LLMs against more than 10 attacks, while baselines fail under unlearning and response manipulations.

As large AI models become increasingly valuable assets, the risk of model weight exfiltration from inference servers grows accordingly. An attacker controlling an inference server may exfiltrate model weights by hiding them within ordinary model outputs, a strategy known as steganography. This work investigates how to verify model responses to defend against such attacks and, more broadly, to detect anomalous or buggy behavior during inference. We formalize model exfiltration as a security game, propose a verification framework that can provably mitigate steganographic exfiltration, and specify the trust assumptions associated with our scheme. To enable verification, we characterize valid sources of non-determinism in large language model inference and introduce two practical estimators for them. We evaluate our detection framework on several open-weight models ranging from 3B to 30B parameters. On MOE-Qwen-30B, our detector reduces exfiltratable information to <0.5% with false-positive rate of 0.01%, corresponding to a >200x slowdown for adversaries. Overall, this work further establishes a foundation for defending against model weight exfiltration and demonstrates that strong protection can be achieved with minimal additional cost to inference providers.

Large Language Model (LLM) deployment requires guiding the LLM to recognize and not answer unsafe prompts while complying with safe prompts. Previous methods for achieving this require adjusting model weights along with other expensive procedures. While recent advances in Sparse Autoencoders (SAEs) have enabled interpretable feature extraction from LLMs, existing approaches lack systematic feature selection methods and principled evaluation of safety-utility tradeoffs. We explored using different steering features and steering strengths using Sparse Auto Encoders (SAEs) to provide a solution. Using an accurate and innovative contrasting prompt method with the AI-Generated Prompts Dataset from teknium/OpenHermes-2p5-Mistral-7B and Air Bench eu-dataset to efficiently choose the best features in the model to steer, we tested this method on Llama-3 8B. We conclude that using this method, our approach achieves an 18.9% improvement in safety performance while simultaneously increasing utility by 11.1%, demonstrating that targeted SAE steering can overcome traditional safety-utility tradeoffs when optimal features are identified through principled selection methods.

Federated Learning (FL) enables decentralized model training across multiple clients without exposing local data, but its distributed feature makes it vulnerable to backdoor attacks. Despite early FL backdoor attacks modifying entire models, recent studies have explored the concept of backdoor-critical (BC) layers, which poison the chosen influential layers to maintain stealthiness while achieving high effectiveness. However, existing BC layers approaches rely on rule-based selection without consideration of the interrelations between layers, making them ineffective and prone to detection by advanced defenses. In this paper, we propose POLAR (POlicy-based LAyerwise Reinforcement learning), the first pipeline to creatively adopt RL to solve the BC layer selection problem in layer-wise backdoor attack. Different from other commonly used RL paradigm, POLAR is lightweight with Bernoulli sampling. POLAR dynamically learns an attack strategy, optimizing layer selection using policy gradient updates based on backdoor success rate (BSR) improvements. To ensure stealthiness, we introduce a regularization constraint that limits the number of modified layers by penalizing large attack footprints. Extensive experiments demonstrate that POLAR outperforms the latest attack methods by up to 40% against six state-of-the-art (SOTA) defenses.

Large Language Models (LLMs) are increasingly vulnerable to a sophisticated form of adversarial prompting known as camouflaged jailbreaking. This method embeds malicious intent within seemingly benign language to evade existing safety mechanisms. Unlike overt attacks, these subtle prompts exploit contextual ambiguity and the flexible nature of language, posing significant challenges to current defense systems. This paper investigates the construction and impact of camouflaged jailbreak prompts, emphasizing their deceptive characteristics and the limitations of traditional keyword-based detection methods. We introduce a novel benchmark dataset, Camouflaged Jailbreak Prompts, containing 500 curated examples (400 harmful and 100 benign prompts) designed to rigorously stress-test LLM safety protocols. In addition, we propose a multi-faceted evaluation framework that measures harmfulness across seven dimensions: Safety Awareness, Technical Feasibility, Implementation Safeguards, Harmful Potential, Educational Value, Content Quality, and Compliance Score. Our findings reveal a stark contrast in LLM behavior: while models demonstrate high safety and content quality with benign inputs, they exhibit a significant decline in performance and safety when confronted with camouflaged jailbreak attempts. This disparity underscores a pervasive vulnerability, highlighting the urgent need for more nuanced and adaptive security strategies to ensure the responsible and robust deployment of LLMs in real-world applications.

Graph Neural Networks (GNNs) have emerged as powerful models for learning from graph-structured data. However, their widespread adoption has raised serious privacy concerns. While prior research has primarily focused on edge-level privacy, a critical yet underexplored threat lies in topology privacy - the confidentiality of the graph's overall structure. In this work, we present a comprehensive study on topology privacy risks in GNNs, revealing their vulnerability to graph-level inference attacks. To this end, we propose a suite of Topology Inference Attacks (TIAs) that can reconstruct the structure of a target training graph using only black-box access to a GNN model. Our findings show that GNNs are highly susceptible to these attacks, and that existing edge-level differential privacy mechanisms are insufficient as they either fail to mitigate the risk or severely compromise model accuracy. To address this challenge, we introduce Private Graph Reconstruction (PGR), a novel defense framework designed to protect topology privacy while maintaining model accuracy. PGR is formulated as a bi-level optimization problem, where a synthetic training graph is iteratively generated using meta-gradients, and the GNN model is concurrently updated based on the evolving graph. Extensive experiments demonstrate that PGR significantly reduces topology leakage with minimal impact on model accuracy. Our code is available at https://github.com/JeffffffFu/PGR.

Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) but pose risks of inadvertently exposing copyrighted or proprietary data, especially when such data is used for training but not intended for distribution. Traditional methods address these leaks only after content is generated, which can lead to the exposure of sensitive information. This study introduces a proactive approach: examining LLMs' internal states before text generation to detect potential leaks. By using a curated dataset of copyrighted materials, we trained a neural network classifier to identify risks, allowing for early intervention by stopping the generation process or altering outputs to prevent disclosure. Integrated with a Retrieval-Augmented Generation (RAG) system, this framework ensures adherence to copyright and licensing requirements while enhancing data privacy and ethical standards. Our results show that analyzing internal states effectively mitigates the risk of copyrighted data leakage, offering a scalable solution that fits smoothly into AI workflows, ensuring compliance with copyright regulations while maintaining high-quality text generation. The implementation is available on GitHub.\footnote{https://github.com/changhu73/Internal_states_leakage}

Detecting machine-generated text (MGT) from contemporary Large Language Models (LLMs) is increasingly crucial amid risks like disinformation and threats to academic integrity. Existing zero-shot detection paradigms, despite their practicality, often exhibit significant deficiencies. Key challenges include: (1) superficial analyses focused on limited textual attributes, and (2) a lack of investigation into consistency across linguistic dimensions such as style, semantics, and logic. To address these challenges, we introduce the \textbf{C}ollaborative \textbf{A}dversarial \textbf{M}ulti-agent \textbf{F}ramework (\textbf{CAMF}), a novel architecture using multiple LLM-based agents. CAMF employs specialized agents in a synergistic three-phase process: \emph{Multi-dimensional Linguistic Feature Extraction}, \emph{Adversarial Consistency Probing}, and \emph{Synthesized Judgment Aggregation}. This structured collaborative-adversarial process enables a deep analysis of subtle, cross-dimensional textual incongruities indicative of non-human origin. Empirical evaluations demonstrate CAMF's significant superiority over state-of-the-art zero-shot MGT detection techniques.

Transformer models have demonstrated exceptional performance and have become indispensable in computer vision (CV) and natural language processing (NLP) tasks. However, recent studies reveal that transformers are susceptible to backdoor attacks. Prior backdoor attack methods typically rely on retraining with clean data or altering the model architecture, both of which can be resource-intensive and intrusive. In this paper, we propose Head-wise Pruning and Malicious Injection (HPMI), a novel retraining-free backdoor attack on transformers that does not alter the model's architecture. Our approach requires only a small subset of the original data and basic knowledge of the model architecture, eliminating the need for retraining the target transformer. Technically, HPMI works by pruning the least important head and injecting a pre-trained malicious head to establish the backdoor. We provide a rigorous theoretical justification demonstrating that the implanted backdoor resists detection and removal by state-of-the-art defense techniques, under reasonable assumptions. Experimental evaluations across multiple datasets further validate the effectiveness of HPMI, showing that it 1) incurs negligible clean accuracy loss, 2) achieves at least 99.55% attack success rate, and 3) bypasses four advanced defense mechanisms. Additionally, relative to state-of-the-art retraining-dependent attacks, HPMI achieves greater concealment and robustness against diverse defense strategies, while maintaining minimal impact on clean accuracy.