Large Language Models (LLMs) have emerged as powerful re-rankers. Recent research has however showed that simple prompt injections embedded within a candidate document (i.e., jailbreak prompt attacks) can significantly alter an LLM's ranking decisions. While this poses serious security risks to LLM-based ranking pipelines, the extent to which this vulnerability persists across diverse LLM families, architectures, and settings remains largely under-explored. In this paper, we present a comprehensive empirical study of jailbreak prompt attacks against LLM rankers. We focus our evaluation on two complementary tasks: (1) Preference Vulnerability Assessment, measuring intrinsic susceptibility via attack success rate (ASR); and (2) Ranking Vulnerability Assessment, quantifying the operational impact on the ranking's quality (nDCG@10). We systematically examine three prevalent ranking paradigms (pairwise, listwise, setwise) under two injection variants: decision objective hijacking and decision criteria hijacking. Beyond reproducing prior findings, we expand the analysis to cover vulnerability scaling across model families, position sensitivity, backbone architectures, and cross-domain robustness. Our results characterize the boundary conditions of these vulnerabilities, revealing critical insights such as that encoder-decoder architectures exhibit strong inherent resilience to jailbreak attacks. We publicly release our code and additional experimental results at https://github.com/ielab/LLM-Ranker-Attack.

Training and fine-tuning large-scale language models largely benefit from collaborative learning, but the approach has been proven vulnerable to gradient inversion attacks (GIAs), which allow adversaries to reconstruct private training data from shared gradients. Existing defenses mainly employ gradient perturbation techniques, e.g., noise injection or gradient pruning, to disrupt GIAs' direct mapping from gradient space to token space. However, these methods often fall short due to the retention of semantics similarity across gradient, embedding, and token spaces. In this work, we propose a novel defense mechanism named GHOST (gradient shield with obfuscated tokens), a token-level obfuscation mechanism that neutralizes GIAs by decoupling the inherent connections across gradient, embedding, and token spaces. GHOST is built upon an important insight: due to the large scale of the token space, there exist semantically distinct yet embedding-proximate tokens that can serve as the shadow substitutes of the original tokens, which enables a semantic disconnection in the token space while preserving the connection in the embedding and gradient spaces. GHOST comprises a searching step, which identifies semantically distinct candidate tokens using a multi-criteria searching process, and a selection step, which selects optimal shadow tokens to ensure minimal disruption to features critical for training by preserving alignment with the internal outputs produced by original tokens. Evaluation across diverse model architectures (from BERT to Llama) and datasets demonstrates the remarkable effectiveness of GHOST in protecting privacy (as low as 1% in recovery rate) and preserving utility (up to 0.92 in classification F1 and 5.45 in perplexity), in both classification and generation tasks against state-of-the-art GIAs and adaptive attack scenarios.

Diffusion models have achieved remarkable progress in image generation, but their increasing deployment raises serious concerns about privacy. In particular, fine-tuned models are highly vulnerable, as they are often fine-tuned on small and private datasets. Membership inference attacks (MIAs) are used to assess privacy risks by determining whether a specific sample was part of a model's training data. Existing MIAs against diffusion models either assume obtaining the intermediate results or require auxiliary datasets for training the shadow model. In this work, we utilized a critical yet overlooked vulnerability: the widely used noise schedules fail to fully eliminate semantic information in the images, resulting in residual semantic signals even at the maximum noise step. We empirically demonstrate that the fine-tuned diffusion model captures hidden correlations between the residual semantics in initial noise and the original images. Building on this insight, we propose a simple yet effective membership inference attack, which injects semantic information into the initial noise and infers membership by analyzing the model's generation result. Extensive experiments demonstrate that the semantic initial noise can strongly reveal membership information, highlighting the vulnerability of diffusion models to MIAs.

LLM-based role-playing has rapidly improved in fidelity, yet stronger adherence to persona constraints commonly increases vulnerability to jailbreak attacks, especially for risky or negative personas. Most prior work mitigates this issue with training-time solutions (e.g., data curation or alignment-oriented regularization). However, these approaches are costly to maintain as personas and attack strategies evolve, can degrade in-character behavior, and are typically infeasible for frontier closed-weight LLMs. We propose a training-free Dual-Cycle Adversarial Self-Evolution framework with two coupled cycles. A Persona-Targeted Attacker Cycle synthesizes progressively stronger jailbreak prompts, while a Role-Playing Defender Cycle distills observed failures into a hierarchical knowledge base of (i) global safety rules, (ii) persona-grounded constraints, and (iii) safe in-character exemplars. At inference time, the Defender retrieves and composes structured knowledge from this hierarchy to guide generation, producing responses that remain faithful to the target persona while satisfying safety constraints. Extensive experiments across multiple proprietary LLMs show consistent gains over strong baselines on both role fidelity and jailbreak resistance, and robust generalization to unseen personas and attack prompts.

Split Learning (SL) offers a framework for collaborative model training that respects data privacy by allowing participants to share the same dataset while maintaining distinct feature sets. However, SL is susceptible to backdoor attacks, in which malicious clients subtly alter their embeddings to insert hidden triggers that compromise the final trained model. To address this vulnerability, we introduce SecureSplit, a defense mechanism tailored to SL. SecureSplit applies a dimensionality transformation strategy to accentuate subtle differences between benign and poisoned embeddings, facilitating their separation. With this enhanced distinction, we develop an adaptive filtering approach that uses a majority-based voting scheme to remove contaminated embeddings while preserving clean ones. Rigorous experiments across four datasets (CIFAR-10, MNIST, CINIC-10, and ImageNette), five backdoor attack scenarios, and seven alternative defenses confirm the effectiveness of SecureSplit under various challenging conditions.

Existing adversarial attacks for VLP models are mostly sample-specific, resulting in substantial computational overhead when scaled to large datasets or new scenarios. To overcome this limitation, we propose Hierarchical Refinement Attack (HRA), a multimodal universal attack framework for VLP models. For the image modality, we refine the optimization path by leveraging a temporal hierarchy of historical and estimated future gradients to avoid local minima and stabilize universal perturbation learning. For the text modality, it hierarchically models textual importance by considering both intra- and inter-sentence contributions to identify globally influential words, which are then used as universal text perturbations. Extensive experiments across various downstream tasks, VLP models, and datasets, demonstrate the superior transferability of the proposed universal multimodal attacks.

This paper reveals that LLM-powered agents exhibit not only demographic bias (e.g., gender, religion) but also intergroup bias under minimal "us" versus "them" cues. When such group boundaries align with the agent-human divide, a new bias risk emerges: agents may treat other AI agents as the ingroup and humans as the outgroup. To examine this risk, we conduct a controlled multi-agent social simulation and find that agents display consistent intergroup bias in an all-agent setting. More critically, this bias persists even in human-facing interactions when agents are uncertain about whether the counterpart is truly human, revealing a belief-dependent fragility in bias suppression toward humans. Motivated by this observation, we identify a new attack surface rooted in identity beliefs and formalize a Belief Poisoning Attack (BPA) that can manipulate agent identity beliefs and induce outgroup bias toward humans. Extensive experiments demonstrate both the prevalence of agent intergroup bias and the severity of BPA across settings, while also showing that our proposed defenses can mitigate the risk. These findings are expected to inform safer agent design and motivate more robust safeguards for human-facing agents.

Vision-language models (VLMs) achieve remarkable performance but remain vulnerable to adversarial attacks. Entropy, a measure of model uncertainty, is strongly correlated with the reliability of VLM. Prior entropy-based attacks maximize uncertainty at all decoding steps, implicitly assuming that every token contributes equally to generation instability. We show instead that a small fraction (about 20%) of high-entropy tokens, i.e., critical decision points in autoregressive generation, disproportionately governs output trajectories. By concentrating adversarial perturbations on these positions, we achieve semantic degradation comparable to global methods while using substantially smaller budgets. More importantly, across multiple representative VLMs, such selective attacks convert 35-49% of benign outputs into harmful ones, exposing a more critical safety risk. Remarkably, these vulnerable high-entropy forks recur across architecturally diverse VLMs, enabling feasible transferability (17-26% harmful rates on unseen targets). Motivated by these findings, we propose Entropy-bank Guided Adversarial attacks (EGA), which achieves competitive attack success rates (93-95%) alongside high harmful conversion, thereby revealing new weaknesses in current VLM safety mechanisms.

Unexploitable example generation aims to transform personal images into their unexploitable (unlearnable) versions before they are uploaded online, thereby preventing unauthorized exploitation of online personal images. Recently, this task has garnered significant research attention due to its critical relevance to personal data privacy. Yet, despite recent progress, existing methods for this task can still suffer from limited practical applicability, as they can fail to generate examples that are broadly unexploitable across different real-world computer vision tasks. To deal with this problem, in this work, we propose a novel Meta Cross-Task Unexploitable Example Generation (MCT-UEG) framework. At the core of our framework, to optimize the unexploitable example generator for effectively producing broadly unexploitable examples, we design a flat-minima-oriented meta training and testing scheme. Extensive experiments show the efficacy of our framework.

Deep neural networks (DNNs) have become valuable intellectual property of model owners, due to the substantial resources required for their development. To protect these assets in the deployed environment, recent research has proposed model usage control mechanisms to ensure models cannot be used without proper authorization. These methods typically lock the utility of the model by embedding an access key into its parameters. However, they often assume static deployment, and largely fail to withstand continual post-deployment model updates, such as fine-tuning or task-specific adaptation. In this paper, we propose ADALOC, to endow key-based model usage control with adaptability during model evolution. It strategically selects a subset of weights as an intrinsic access key, which enables all model updates to be confined to this key throughout the evolution lifecycle. ADALOC enables using the access key to restore the keyed model to the latest authorized states without redistributing the entire network (i.e., adaptation), and frees the model owner from full re-keying after each model update (i.e., lock preservation). We establish a formal foundation to underpin ADALOC, providing crucial bounds such as the errors introduced by updates restricted to the access key. Experiments on standard benchmarks, such as CIFAR-100, Caltech-256, and Flowers-102, and modern architectures, including ResNet, DenseNet, and ConvNeXt, demonstrate that ADALOC achieves high accuracy under significant updates while retaining robust protections. Specifically, authorized usages consistently achieve strong task-specific performance, while unauthorized usage accuracy drops to near-random guessing levels (e.g., 1.01% on CIFAR-100), compared to up to 87.01% without ADALOC. This shows that ADALOC can offer a practical solution for adaptive and protected DNN deployment in evolving real-world scenarios.

Recent AI regulations call for data that remain useful for innovation while resistant to misuse, balancing utility with protection at the model level. Existing approaches either perturb data to make it unlearnable or retrain models to suppress transfer, but neither governs inference by unknown models, and both typically require control over training. We propose non-transferable examples (NEs), a training-free and data-agnostic input-side usage-control mechanism. We recode inputs within a model-specific low-sensitivity subspace, preserving outputs for the authorized model while reducing performance on unauthorized models through subspace misalignment. We establish formal bounds that guarantee utility for the authorized model and quantify deviation for unauthorized ones, with the Hoffman-Wielandt inequality linking degradation to spectral differences. Empirically, NEs retain performance on diverse vision backbones and state-of-the-art vision-language models under common preprocessing, whereas non-target models collapse even with reconstruction attempts. These results establish NEs as a practical means to preserve intended data utility while preventing unauthorized exploitation. Our project is available at https://trusted-system-lab.github.io/model-specificity

Large language models (LLMs) frequently memorize sensitive or personal information, raising significant privacy concerns. Existing variants of differential privacy stochastic gradient descent (DPSGD) inject uniform noise into every gradient step, significantly extending training time and reducing model accuracy. We propose that concentrating noise primarily on gradients associated with sensitive tokens can substantially decrease DP training time, strengthen the protection of sensitive information, and simultaneously preserve the model's performance on non-sensitive data. We operationalize this insight through Adaptive Token-Weighted Differential Privacy (ATDP), a modification of vanilla DP-SGD that adaptively assigns different gradient weights to sensitive and non-sensitive tokens. By employing a larger noise scale at the early stage of training, ATDP rapidly disrupts memorization of sensitive content. As a result, ATDP only requires a few additional epochs of lightweight post-processing following standard fine-tuning, injecting targeted noise primarily on parameters corresponding to sensitive tokens, thus minimally affecting the model's general capabilities. ATDP can be seamlessly integrated into any existing DP-based fine-tuning pipeline or directly applied to non-private models as a fast privacy-enhancing measure. Additionally, combined with an initial redacted fine-tuning phase, ATDP forms a streamlined DP pipeline that achieves comparable canary protection to state-of-the-art DP-SGD methods, significantly reduces the computational overhead of DP fine-tuning, shortening training time by approximately 90 percent, while achieving comparable or superior privacy protection and minimal accuracy degradation.

The widespread distribution of Large Language Models (LLMs) through public platforms like Hugging Face introduces significant security challenges. While these platforms perform basic security scans, they often fail to detect subtle manipulations within the embedding layer. This work identifies a novel class of deployment phase attacks that exploit this vulnerability by injecting imperceptible perturbations directly into the embedding layer outputs without modifying model weights or input text. These perturbations, though statistically benign, systematically bypass safety alignment mechanisms and induce harmful behaviors during inference. We propose Search based Embedding Poisoning(SEP), a practical, model agnostic framework that introduces carefully optimized perturbations into embeddings associated with high risk tokens. SEP leverages a predictable linear transition in model responses, from refusal to harmful output to semantic deviation to identify a narrow perturbation window that evades alignment safeguards. Evaluated across six aligned LLMs, SEP achieves an average attack success rate of 96.43% while preserving benign task performance and evading conventional detection mechanisms. Our findings reveal a critical oversight in deployment security and emphasize the urgent need for embedding level integrity checks in future LLM defense strategies.