Vision-language models (VLMs) are vulnerable to adversarial image perturbations. Existing works based on adversarial training against task-specific adversarial examples are computationally expensive and often fail to generalize to unseen attack types. To address these limitations, we introduce Paraphrase-Decomposition-Aggregation (PDA), a training-free defense framework that leverages text augmentation to enhance VLM robustness under diverse adversarial image attacks. PDA performs prompt paraphrasing, question decomposition, and consistency aggregation entirely at test time, thus requiring no modification on the underlying models. To balance robustness and efficiency, we instantiate PDA as invariants that reduce the inference cost while retaining most of its robustness gains. Experiments on multiple VLM architectures and benchmarks for visual question answering, classification, and captioning show that PDA achieves consistent robustness gains against various adversarial perturbations while maintaining competitive clean accuracy, establishing a generic, strong and practical defense framework for VLMs during inference.

Large language models (LLMs) are increasingly deployed on edge devices under strict computation and quantization constraints, yet their security implications remain unclear. We study query-based knowledge extraction from quantized edge-deployed LLMs under realistic query budgets and show that, although quantization introduces noise, it does not remove the underlying semantic knowledge, allowing substantial behavioral recovery through carefully designed queries. To systematically analyze this risk, we propose \textbf{CLIQ} (\textbf{Cl}ustered \textbf{I}nstruction \textbf{Q}uerying), a structured query construction framework that improves semantic coverage while reducing redundancy. Experiments on quantized Qwen models (INT8/INT4) demonstrate that CLIQ consistently outperforms original queries across BERTScore, BLEU, and ROUGE, enabling more efficient extraction under limited budgets. These results indicate that quantization alone does not provide effective protection against query-based extraction, highlighting a previously underexplored security risk in edge-deployed LLMs.

Speech deepfake source verification systems aims to determine whether two synthetic speech utterances originate from the same source generator, often assuming that the resulting source embeddings are independent of speaker traits. However, this assumption remains unverified. In this paper, we first investigate the impact of speaker factors on source verification. We propose a speaker-disentangled metric learning (SDML) framework incorporating two novel loss functions. The first leverages Chebyshev polynomial to mitigate gradient instability during disentanglement optimization. The second projects source and speaker embeddings into hyperbolic space, leveraging Riemannian metric distances to reduce speaker information and learn more discriminative source features. Experimental results on MLAAD benchmark, evaluated under four newly proposed protocols designed for source-speaker disentanglement scenarios, demonstrate the effectiveness of SDML framework. The code, evaluation protocols and demo website are available at https://github.com/xxuan-acoustics/RiemannSD-Net.

Although text-to-image diffusion models exhibit remarkable generative power, concept erasure techniques are essential for their safe deployment to prevent the creation of harmful content. This has fostered a dynamic interplay between the development of erasure defenses and the adversarial probes designed to bypass them, and this co-evolution has progressively enhanced the efficacy of erasure methods. However, this adversarial co-evolution has converged on a narrow, text-centric paradigm that equates erasure with severing the text-to-image mapping, ignoring that the underlying visual knowledge related to undesired concepts still persist. To substantiate this claim, we investigate from a visual perspective, leveraging DDIM inversion to probe whether a generative pathway for the erased concept can still be found. However, identifying such a visual generative pathway is challenging because standard text-guided DDIM inversion is actively resisted by text-centric defenses within the erased model. To address this, we introduce TINA, a novel Text-free INversion Attack, which enforces this visual-only probe by operating under a null-text condition, thereby avoiding existing text-centric defenses. Moreover, TINA integrates an optimization procedure to overcome the accumulating approximation errors that arise when standard inversion operates without its usual textual guidance. Our experiments demonstrate that TINA regenerates erased concepts from models treated with state-of-the-art unlearning. The success of TINA proves that current methods merely obscure concepts, highlighting an urgent need for paradigms that operate directly on internal visual knowledge.

Federated Learning (FL) enables collaborative model training by sharing model updates instead of raw data, aiming to protect user privacy. However, recent studies reveal that these shared updates can inadvertently leak sensitive training data through gradient inversion attacks (GIAs). Among them, active GIAs are particularly powerful, enabling high-fidelity reconstruction of individual samples even under large batch sizes. Nevertheless, existing approaches often require architectural modifications, which limit their practical applicability. In this work, we bridge this gap by introducing the Activation REcovery via Sparse inversion (ARES) attack, an active GIA designed to reconstruct training samples from large training batches without requiring architectural modifications. Specifically, we formulate the recovery problem as a noisy sparse recovery task and solve it using the generalized Least Absolute Shrinkage and Selection Operator (Lasso). To extend the attack to multi-sample recovery, ARES incorporates the imprint method to disentangle activations, enabling scalable per-sample reconstruction. We further establish the expected recovery rate and derive an upper bound on the reconstruction error, providing theoretical guarantees for the ARES attack. Extensive experiments on CNNs and MLPs demonstrate that ARES achieves high-fidelity reconstruction across diverse datasets, significantly outperforming prior GIAs under large batch sizes and realistic FL settings. Our results highlight that intermediate activations pose a serious and underestimated privacy risk in FL, underscoring the urgent need for stronger defenses.

LLM agents are highly vulnerable to Indirect Prompt Injection (IPI), where adversaries embed malicious directives in untrusted tool outputs to hijack execution. Most existing defenses treat IPI as an input-level semantic discrimination problem, which often fails to generalize to unseen payloads. We propose a new paradigm, action-level causal attribution, which secures agents by asking why a particular tool call is produced. The central goal is to distinguish tool calls supported by the user's intent from those causally driven by untrusted observations. We instantiate this paradigm with AttriGuard, a runtime defense based on parallel counterfactual tests. For each proposed tool call, AttriGuard verifies its necessity by re-executing the agent under a control-attenuated view of external observations. Technically, AttriGuard combines teacher-forced shadow replay to prevent attribution confounding, hierarchical control attenuation to suppress diverse control channels while preserving task-relevant information, and a fuzzy survival criterion that is robust to LLM stochasticity. Across four LLMs and two agent benchmarks, AttriGuard achieves 0% ASR under static attacks with negligible utility loss and moderate overhead. Importantly, it remains resilient under adaptive optimization-based attacks in settings where leading defenses degrade significantly.

Recent advances in generative models have amplified the risk of malicious misuse of speech synthesis technologies, enabling adversaries to impersonate target speakers and access sensitive resources. Although speech deepfake detection has progressed rapidly, most existing countermeasures lack formal robustness guarantees or fail to generalize to unseen generation techniques. We propose PV-VASM, a probabilistic framework for verifying the robustness of voice anti-spoofing models (VASMs). PV-VASM estimates the probability of misclassification under text-to-speech (TTS), voice cloning (VC), and parametric signal transformations. The approach is model-agnostic and enables robustness verification against unseen speech synthesis techniques and input perturbations. We derive a theoretical upper bound on the error probability and validate the method across diverse experimental settings, demonstrating its effectiveness as a practical robustness verification tool.

The Model Context Protocol (MCP) is an open and standardized interface that enables large language models (LLMs) to interact with external tools and services, and is increasingly adopted by AI agents. However, the security of MCP-based systems remains largely unexplored.In this work, we conduct a large-scale security analysis of MCP servers integrated within MCP clients. We show that treating MCP servers as trusted entities without authenticating the caller identity is fundamentally insecure. Since MCP servers often cannot distinguish who is invoking a request, a single authorization decision may implicitly grant access to multiple, potentially untrusted callers.Our empirical study reveals that most MCP servers rely on persistent authorization states, allowing tool invocations after an initial authorization without re-authentication, regardless of the caller. In addition, many MCP servers fail to enforce authentication at the per-tool level, enabling unauthorized access to sensitive operations.These findings demonstrate that one-time authorization and server-level trust significantly expand the attack surface of MCP-based systems, highlighting the need for explicit caller authentication and fine-grained authorization mechanisms.

Multi-tenant LLM serving frameworks widely adopt shared Key-Value caches to enhance efficiency. However, this creates side-channel vulnerabilities enabling prompt leakage attacks. Prior studies identified these attack surfaces yet focused on expanding attack vectors rather than optimizing attack performance, reporting impractically high attack costs that underestimate the true privacy risk. We propose OptiLeak, a reinforcement learning-enhanced framework that maximizes prompt reconstruction efficiency through two-stage fine-tuning. Our key insight is that domain-specific ``hard tokens'' -- terms difficult to predict yet carrying sensitive information -- can be automatically identified via likelihood ranking and used to construct preference pairs for Direct Preference Optimization, eliminating manual annotation. This enables effective preference alignment while avoiding the overfitting issues of extended supervised fine-tuning. Evaluated on three benchmarks spanning medical and financial domains, OptiLeak achieves up to $12.48\times$ reduction in average requests per token compared to baseline approaches, with consistent improvements across model scales from 3B to 14B parameters. Our findings demonstrate that cache-based prompt leakage poses a more severe threat than previously reported, underscoring the need for robust cache isolation in production deployments.

The proliferation of AI-generated content has facilitated sophisticated face manipulation, severely undermining visual integrity and posing unprecedented challenges to intellectual property. In response, a common proactive defense leverages fragile watermarks to detect, localize, or even recover manipulated regions. However, these methods always assume an adversary unaware of the embedded watermark, overlooking their inherent vulnerability to watermark removal attacks. Furthermore, this fragility is exacerbated in the commonly used dual-watermark strategy that adds a robust watermark for image ownership verification, where mutual interference and limited embedding capacity reduce the fragile watermark's effectiveness. To address the gap, we propose RecoverMark, a watermarking framework that achieves robust manipulation localization, content recovery, and ownership verification simultaneously. Our key insight is twofold. First, we exploit a critical real-world constraint: an adversary must preserve the background's semantic consistency to avoid visual detection, even if they apply global, imperceptible watermark removal attacks. Second, using the image's own content (face, in this paper) as the watermark enhances extraction robustness. Based on these insights, RecoverMark treats the protected face content itself as the watermark and embeds it into the surrounding background. By designing a robust two-stage training paradigm with carefully crafted distortion layers that simulate comprehensive potential attacks and a progressive training strategy, RecoverMark achieves a robust watermark embedding in no fragile manner for image manipulation localization, recovery, and image IP protection simultaneously. Extensive experiments demonstrate the proposed RecoverMark's robustness against both seen and unseen attacks and its generalizability to in-distribution and out-of-distribution data.

Large Vision-Language Models (LVLMs) are foundational to modern multimodal applications, yet their susceptibility to adversarial attacks remains a critical concern. Prior white-box attacks rarely generalize across tasks, and black-box methods depend on expensive transfer, which limits efficiency. The vision encoder, standardized and often shared across LVLMs, provides a stable gray-box pivot with strong cross-model transfer. Building on this premise, we introduce PA-Attack (Prototype-Anchored Attentive Attack). PA-Attack begins with a prototype-anchored guidance that provides a stable attack direction towards a general and dissimilar prototype, tackling the attribute-restricted issue and limited task generalization of vanilla attacks. Building on this, we propose a two-stage attention enhancement mechanism: (i) leverage token-level attention scores to concentrate perturbations on critical visual tokens, and (ii) adaptively recalibrate attention weights to track the evolving attention during the adversarial process. Extensive experiments across diverse downstream tasks and LVLM architectures show that PA-Attack achieves an average 75.1% score reduction rate (SRR), demonstrating strong attack effectiveness, efficiency, and task generalization in LVLMs. Code is available at https://github.com/hefeimei06/PA-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.

Retrieval-augmented generation (RAG) is increasingly deployed in real-world applications, where its reference-grounded design makes outputs appear trustworthy. This trust has spurred research on poisoning attacks that craft malicious content, inject it into knowledge sources, and manipulate RAG responses. However, when evaluated in practical RAG systems, existing attacks suffer from severely degraded effectiveness. This gap stems from two overlooked realities: (i) content is often processed before use, which can fragment the poison and weaken its effect, and (ii) users often do not issue the exact queries anticipated during attack design. These factors can lead practitioners to underestimate risks and develop a false sense of security. To better characterize the threat to practical systems, we present Confundo, a learning-to-poison framework that fine-tunes a large language model as a poison generator to achieve high effectiveness, robustness, and stealthiness. Confundo provides a unified framework supporting multiple attack objectives, demonstrated by manipulating factual correctness, inducing biased opinions, and triggering hallucinations. By addressing these overlooked challenges, Confundo consistently outperforms a wide range of purpose-built attacks across datasets and RAG configurations by large margins, even in the presence of defenses. Beyond exposing vulnerabilities, we also present a defensive use case that protects web content from unauthorized incorporation into RAG systems via scraping, with no impact on user experience.

With the deep integration of facial recognition into online banking, identity verification, and other networked services, achieving effective decoupling of identity information from visual representations during image storage and transmission has become a critical challenge for privacy protection. To address this issue, we propose SIDeR, a Semantic decoupling-driven framework for unrestricted face privacy protection. SIDeR decomposes a facial image into a machine-recognizable identity feature vector and a visually perceptible semantic appearance component. By leveraging semantic-guided recomposition in the latent space of a diffusion model, it generates visually anonymous adversarial faces while maintaining machine-level identity consistency. The framework incorporates momentum-driven unrestricted perturbation optimization and a semantic-visual balancing factor to synthesize multiple visually diverse, highly natural adversarial samples. Furthermore, for authorized access, the protected image can be restored to its original form when the correct password is provided. Extensive experiments on the CelebA-HQ and FFHQ datasets demonstrate that SIDeR achieves a 99% attack success rate in black-box scenarios and outperforms baseline methods by 41.28% in PSNR-based restoration quality.

As realistic AI-generated images threaten digital authenticity, we address the generalization failure of generative artifact-based detectors by exploiting the intrinsic properties of the camera imaging pipeline. Concretely, we investigate color correlations induced by the color filter array (CFA) and demosaicing, and propose a Demosaicing-guided Color Correlation Training (DCCT) framework for AI-generated image detection. By simulating the CFA sampling pattern, we decompose each color image into a single-channel input (as the condition) and the remaining two channels as the ground-truth targets (for prediction). A self-supervised U-Net is trained to model the conditional distribution of the missing channels from the given one, parameterized via a mixture of logistic functions. Our theoretical analysis reveals that DCCT targets a provable distributional difference in color-correlation features between photographic and AI-generated images. By leveraging these distinct features to construct a binary classifier, DCCT achieves state-of-the-art generalization and robustness, significantly outperforming prior methods across over 20 unseen generators.

Autonomous code agents built on large language models are reshaping software and AI development through tool use, long-horizon reasoning, and self-directed interaction. However, this autonomy introduces a previously unrecognized security risk: agentic interaction fundamentally expands the LLM attack surface, enabling systematic probing and recovery of hidden system prompts that guide model behavior. We identify system prompt extraction as an emergent vulnerability intrinsic to code agents and present \textbf{\textsc{JustAsk}}, a self-evolving framework that autonomously discovers effective extraction strategies through interaction alone. Unlike prior prompt-engineering or dataset-based attacks, \textsc{JustAsk} requires no handcrafted prompts, labeled supervision, or privileged access beyond standard user interaction. It formulates extraction as an online exploration problem, using Upper Confidence Bound-based strategy selection and a hierarchical skill space spanning atomic probes and high-level orchestration. These skills exploit imperfect system-instruction generalization and inherent tensions between helpfulness and safety. Evaluated on \textbf{41} black-box commercial models across multiple providers, \textsc{JustAsk} consistently achieves full or near-complete system prompt recovery, revealing recurring design- and architecture-level vulnerabilities. Our results expose system prompts as a critical yet largely unprotected attack surface in modern agent systems.

Adapter-based Federated Large Language Models (FedLLMs) are widely adopted to reduce the computational, storage, and communication overhead of full-parameter fine-tuning for web-scale applications while preserving user privacy. By freezing the backbone and training only compact low-rank adapters, these methods appear to limit gradient leakage and thwart existing Gradient Inversion Attacks (GIAs). Contrary to this assumption, we show that low-rank adapters create new, exploitable leakage channels. We propose the Unordered-word-bag-based Text Reconstruction (UTR) attack, a novel GIA tailored to the unique structure of adapter-based FedLLMs. UTR overcomes three core challenges: low-dimensional gradients, frozen backbones, and combinatorially large reconstruction spaces by: (i) inferring token presence from attention patterns in frozen layers, (ii) performing sentence-level inversion within the low-rank subspace of adapter gradients, and (iii) enforcing semantic coherence through constrained greedy decoding guided by language priors. Extensive experiments across diverse models (GPT2-Large, BERT, Qwen2.5-7B) and datasets (CoLA, SST-2, Rotten Tomatoes) demonstrate that UTR achieves near-perfect reconstruction accuracy (ROUGE-1/2 > 99), even with large batch size settings where prior GIAs fail completely. Our results reveal a fundamental tension between parameter efficiency and privacy in FedLLMs, challenging the prevailing belief that lightweight adaptation inherently enhances security. Our code and data are available at https://github.com/shwksnshwowk-wq/GIA.

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.

Box-free model watermarking has gained significant attention in deep neural network (DNN) intellectual property protection due to its model-agnostic nature and its ability to flexibly manage high-entropy image outputs from generative models. Typically operating in a black-box manner, it employs an encoder-decoder framework for watermark embedding and extraction. While existing research has focused primarily on the encoders for the robustness to resist various attacks, the decoders have been largely overlooked, leading to attacks against the watermark. In this paper, we identify one such attack against the decoder, where query responses are utilized to obtain backpropagated gradients to train a watermark remover. To address this issue, we propose Decoder Gradient Shields (DGSs), a family of defense mechanisms, including DGS at the output (DGS-O), at the input (DGS-I), and in the layers (DGS-L) of the decoder, with a closed-form solution for DGS-O and provable performance for all DGS. Leveraging the joint design of reorienting and rescaling of the gradients from watermark channel gradient leaking queries, the proposed DGSs effectively prevent the watermark remover from achieving training convergence to the desired low-loss value, while preserving image quality of the decoder output. We demonstrate the effectiveness of our proposed DGSs in diverse application scenarios. Our experimental results on deraining and image generation tasks with the state-of-the-art box-free watermarking show that our DGSs achieve a defense success rate of 100% under all settings.

As LLMs gain persuasive agentic capabilities through extended dialogues, they introduce novel risks in multi-turn conversational scams that single-turn safety evaluations fail to capture. We systematically study these risks using a controlled LLM-to-LLM simulation framework across multi-turn scam scenarios. Evaluating eight state-of-the-art models in English and Chinese, we analyze dialogue outcomes and qualitatively annotate attacker strategies, defensive responses, and failure modes. Results reveal that scam interactions follow recurrent escalation patterns, while defenses employ verification and delay mechanisms. Furthermore, interactional failures frequently stem from safety guardrail activation and role instability. Our findings highlight multi-turn interactional safety as a critical, distinct dimension of LLM behavior.