Retrieval-Augmented Generation (RAG) grounds large language models in external corpora, but implicit trust in retrieved documents creates a critical attack surface: PoisonedRAG shows that a handful of crafted passages can dominate dense retrieval and steer generation toward attacker-chosen answers. We present the Tri-Layer Sieve, a middleware defense that sanitizes retrieved evidence through cross-embedding-space clustering with an independent judge model, structural filtering of trigger-payload artifacts, and LLM consistency verification. The design exploits a key weakness of retrieval-stage poisoning: a single document must satisfy one embedding geometry, one internal Trigger-Payload structure, and one generation objective - rarely all three simultaneously, a fragility that persists even against an adaptive attacker who paraphrases around it. On Natural Questions, HotpotQA, and MS-MARCO with Contriever retrieval (k=50), the Sieve reduces black-box Attack Success Rate from 67.0/87.0/64.0% to 3.0/14.0/4.0%, mitigates white-box HotFlip attacks from ~74% to 27.8% on NQ with Layer 3 enabled, and drives poisoned-document MRR to 0.000, while restoring clean accuracy from 13-33% under attack to 58-76%. Under an architecture-aware adversary who paraphrases triggers to evade the structural filter, enabling the consistency layer halves adaptive ASR (32.0% to 15.0% on NQ) while raising clean accuracy by 18 points, at an added latency of ~16-19 s/query under live retrieval.
The rapid development of generative portrait models has raised growing concerns about privacy leakage and identity misuse. In particular, audio-driven 3D talking face generation can reconstruct a reusable 3D portrait of a target person from a monocular video and animate it with arbitrary speech, making realistic identity impersonation alarmingly practical. Existing proactive defenses mainly operate in the visual domain by injecting subtle perturbations into acial regions to disrupt identity acquisition. However, such perturbations often compromise visual quality due to the strong structural priors and social sensitivity of human faces, and are easily weakened by common real-world transformations such as resizing. To overcome these limitations, we propose an imperceptible audio defense for audio-driven 3D talking face generation by shifting protection from the visual modality to the audio modality. Specifically,we exploit psychoacoustic masking to hide protective perturbations within perceptually masked frequency regions of the speech signal, thereby reducing perceptual distortion while suppressing reliable facial animation. Extensive experiments demonstrate that the proposed method effectively degrades 3D talking face generation while preserving favorable perceptual quality. These findings highlight psychoacoustically guided audio perturbations as a practical and promising direction for privacy-preserving portrait protection.
Multi-norm adversarial defense aims to protect neural networks against perturbations defined by different norm constraints, but existing methods typically optimize competing robustness objectives within a single parameter configuration, leading to substantial training cost and unfavorable robustness trade-offs. We propose Robust CurveMoE, an efficient mixture-of-experts framework that connects models specialized for different perturbation norms through a low-loss path and exploits the complementary robustness profiles of models along this path. Robust CurveMoE derives clean and norm-specialized experts from robustness-constrained curve locations and selectively expertizes only influential layers, while sharing the remaining parameters across routing paths. To further reduce curve-construction cost, we introduce contribution-guided partial updating, which selects influential curve parameters using initialization-based gradient scores. We also theoretically bound the objective gap between partial and full curve optimization. Experiments on CIFAR-100 and ImageNet-100 with WideResNet and Vision Transformer architectures show that Robust CurveMoE consistently improves clean, norm-specific, and Union accuracy over MSD and ERMC. In particular, it improves Union accuracy by 2.37 and 2.13 percentage points over the strongest baseline on CIFAR-100 and ImageNet-100, respectively. Extensive ablations further validate the effectiveness of partial updating, selective expertization, and robustness-constrained expert selection.
Retrieval-augmented generation (RAG) improves the factuality of large language models by grounding responses in external documents, but it also exposes a critical security vulnerability: adversarial documents injected into the knowledge database can enter the context window and steer the model toward targeted incorrect answers. Existing post-retrieval defenses rely on instruction following, parametric knowledge, or text-level consistency, all of which can be imitated or optimized against by adaptive attackers. We propose RAGSentinel, a training-free, label-free defense for black-box RAG systems. RAGSentinel uses a surrogate encoder to measure query-conditioned hidden-state shifts induced by retrieved documents, removes shared topic directions, and filters poisoned documents as geometric outliers from a robust majority consensus. We prove that, under an honest-majority assumption and a representation-level separation condition, RAGSentinel exactly recovers a poison-free majority-sized context. Experiments across three question-answering datasets, three LLM families, and multiple poisoning attacks show that RAGSentinel consistently achieves low attack success rates while preserving competitive accuracy and remaining effective against adaptive attacks with full pipeline knowledge.
Tong Zhang, Motasem Alfarra, Carlos Hinojosa +2cs.AI
As text-to-image generative models advance, they raise critical safety concerns, particularly the generation of Not-Safe-For-Work (NSFW) content such as violence and nudity, further exacerbated by red-teaming adversarial attacks. Existing defenses predominantly operate under white-box assumptions, relying on text encoder optimization, weight editing, or inference-time intervention, and fundamentally cannot scale to proprietary models. Black-box alternatives based on LLM prompt rewriting offer broader applicability, yet fail in a critical regime we identify as the \textit{benign adversarial} problem: prompts that are linguistically safe but still trigger harmful generation due to the model's learned data distribution. We propose DiSCO, a zero-shot, strictly black-box defense that operates entirely at the prompt level as a plug-and-play module, requiring no model retraining, fine-tuning, or access to model internals. DiSCO performs distribution-guided suffix expansion via beam search, optimized through contrastive scoring over safe and unsafe image pools generated by the target model itself, with iterative adaptive feedback until safe content is produced. We demonstrate that DiSCO consistently enhances the safety of both undefended and defended models on the I2P benchmark under multiple red-teaming attacks, achieving 37.7% and 25.13% ASR reduction, respectively, while maintaining semantic fidelity and improving image coherence. As a black-box, architecture-agnostic module, DiSCO can be readily applied to any text-to-image system without necessitating any changes to the model itself.
Multimodal Retrieval Augmented Generation (M-RAG) is increasingly vulnerable to adversarial attacks where malicious data are crafted to produce embeddings that align with benign entries in the vector space, deceiving retrieval and inducing harmful outputs. Existing defenses primarily operate at query time, relying on auxiliary detectors, similarity re-ranking, or feature-consistency checks. However, these approaches suffer from non-trivial inference overhead, generalize poorly to unseen attack strategies, and often assume specific attack distributions. To address this, we propose DSPrompt, a Dynamic Soft Prompt defense framework that directly reshapes the retriever's embedding semantics, without modifying the retrieval pipeline. It inserts few learnable soft prompts into each layer of the visual and textual encoders of a frozen retriever, utilizing a shallow-to-deep length schedule that is adaptive to the capacity in the model layers. These prompts are trained under a dynamic min-max scheme: an online multimodal attacker continually crafts hard adversarial documents against the current retriever, while the defender is updated to push such documents out of the top-k while preserving the ranking and diversity of benign evidence. Because the defended encoder can be pre-computed and indexed exactly as in standard dense retrieval, DSPrompt incurs no additional per-query optimization and introduces fewer than 1% additional parameters. Extensive experiments across four benchmarks and three representative poisoning attacks show that DSPrompt substantially reduces the attack success rate and poison retrieval rate while maintaining near-lossless retrieval utility and generation fidelity, consistently outperforming existing defense baselines at a fraction of their computational cost.
Retrieval-augmented generation (RAG) answers a question by retrieving passages from a vector store and trusting them as context, so anyone who can add documents can try to steer the answer. A recent, appealing defense filters poisoning at ingestion, rejecting any document that behaves like a hub. We show it -- and every ingestion-time filter -- is defeated by a coordinated adversary that injects a handful of individually unremarkable documents which together surround one target query and seize its top-k (on BGE-large / BEIR, m=10 documents take 10/10; 9.9/10 on a live HNSW index). The attack is not theoretical. Realized as ordinary fluent text and run end-to-end through a BGE-large + HNSW + Qwen2.5-7B pipeline, it makes the generator emit the attacker's planted claim in 88% of targets, versus 0% without the injection. And no admission-time defense stops it: at ingestion an attack cone is geometrically identical to a legitimate niche upload, so -- measuring this directly -- the strongest trained classifier, given every feature and thousands of examples, separates the two no better than chance, catching 4.2% of attacks at a 1% false-positive rate. We prove this limit for the entire class of ingestion-time statistics (any decision from documents and reference queries alone), and it reproduces -- and worsens -- across two corpora and five encoders. The one signal that separates an attack from legitimate niche ingestion -- a query's demand -- is invisible before retrieval, which is also the escape: a retrieval-time detector that observes demand catches 100% of the attacks at the same 1% false-positive rate. Coverage of the query space by an admission gate is not containment of coordinated poisoning; robust defense must move past the front door, to demand.
Modern image editors combine vision-language models (VLMs) with diffusion transformer backbones to modify a single reference image according to instructions without fine-tuning. This capability also enables unauthorized manipulation of publicly released images. Existing inference-time defenses either invalidate edits through conspicuous corruption, thereby exposing the protection, or allow them to proceed with identity or reference content drift, thereby failing to prevent the editing behavior itself. We instead target a stealthy and harmless no-op in which the requested edit is suppressed, the output remains natural and source-preserving without conspicuous artifacts or identity replacement, and harmful semantics requested by malicious instructions are absent. We propose NullEdit, which targets the VLM representation jointly formed from the reference image and instruction before it conditions the downstream DiT backbone. Using normal-edit and no-edit anchors, NullEdit redirects this representation, while cross-prompt gradient averaging transfers protection to held out instructions. Across Step1X-Edit and Qwen-Image-Edit on CelebA-HQ and VGGFace2, NullEdit reduces the EditReward IF score by 0.813 on average relative to the SOTA baseline while preserving subject identity and source content.
Jonas Grebe, Hossein Shakibania, Tobias Braun +2cs.CV
The rise of powerful, accessible image-editing models such as FLUX.2 has brought high-fidelity editing within broad reach. Their capabilities now extend beyond localized modifications to extracting and recontextualizing objects and identities in entirely new scenes. By allowing prompt and generation tokens to attend directly to reference-image tokens, modern models blur the boundary between conventional editing and text-to-image synthesis. This expanded generative freedom also broadens the space of potential misuse, as harmful transformations are no longer confined to a predictable set of localized edits. Existing anti-edit defenses are designed to disrupt the semantic bottleneck of the reference-image encoding in legacy diffusion pipelines. However, newer editors distill reference information through joint-attention blocks, thereby often circumventing these protections. We therefore introduce VETO, a subtle anti-edit cloak that disrupts this inner mechanism through which modern models read the source image. Additionally, as existing editing benchmarks leave comprehensive recontextualizations largely untested, we introduce VetoBench, which evaluates defenses not only on conventional localized edits but also on broader contextual shifts. Across two contemporary editing models and three benchmarks, VETO consistently outperforms existing defenses while providing a stronger protection-fidelity trade-off.
Nikos Athanasiou, Ilya A. Petrov, Angela Yao +7cs.CV
Vision and vision-language models rely on high-level visual representations that are increasingly used across recognition, retrieval, and multimodal reasoning pipelines. However, recent advances in generative modeling have shown that such features can often be inverted, enabling realistic reconstructions of the underlying image and raising significant privacy risks. We revisit this problem through the lens of reconstruction and propose TrustCLIP, a reconstruction-driven framework that treats a feature-conditioned generator as an explicit privacy adversary. TrustCLIP learns a projection between encoder features and downstream modules that is explicitly optimized to degrade the reconstructions produced by generative attackers while retaining the necessary signals for downstream tasks. Unlike prior defenses that rely on discriminative privacy metrics, TrustCLIP directly optimizes against a generative reconstruction attacker, targeting a threat not captured by standard evaluation protocols. We demonstrate its effectiveness in both conventional classification and multimodal large language model pipelines. Across these settings, TrustCLIP consistently reduces the fidelity of generative inversions while maintaining downstream task performance. Project page: https://atnikos.github.io/trustclip/
Jungkon Kim, Cheolseung Jung, Jong-Min Choi +1cs.CV
Face-swapping deepfakes pose an escalating threat to personal privacy by enabling unauthorized identity manipulation. While adversarial approaches have demonstrated success against black-box face recognition (FR) models, their applicability to face-swapping scenarios remains underexplored. In particular, reliance on fixed or random targets yields ambiguous latent guidance, and the lack of explicit spatial constraints causes perturbations to spill into identity-irrelevant regions. These issues are further exacerbated by identity-style disentanglement, which suppresses adversarial signals during deepfake generation. In this paper, we present Phantom, a unified face-swap deepfake protection framework that jointly constrains perturbations in latent and spatial domains. Phantom adaptively synthesizes identity-shifted yet attribute-preserving targets to guide identity-aware latent optimization, and applies masked perturbations confined to semantically relevant facial regions. Extensive experiments on state-of-the-art face-swapping deepfakes demonstrate that Phantom improves protection success rates in dodging scenarios by 27.8%, 25.6%, and 16.6% on UniFace, INSwapper, and SimSwap, respectively, while also enhancing visual quality. Furthermore, Phantom generalizes to impersonation scenario, yielding up to 10.2% higher protection while improving perceptual fidelity. These results underscore the effectiveness of jointly leveraging latent and spatial constraints for robust and coherent facial privacy protection.
Neural networks are vulnerable to meticulously crafted adversarial examples, leading to high-confidence misclassifications in image classification tasks. Due to their consistency with regular input patterns and the absence of reliance on the target model and its output information, transferable adversarial attacks exhibit a notably high stealthiness and detection difficulty, making them a significant focus of defense. In this work, we propose a deep learning defense known as multi-source adversarial perturbations elimination (MAPE) to counter diverse transferable attacks. MAPE comprises the single-source adversarial perturbation elimination (SAPE) mechanism and the pre-trained models probabilistic scheduling algorithm (PPSA). SAPE utilizes a thoughtfully designed channel-attention U-Net as the defense model and employs adversarial examples generated by a pre-trained model (e.g., ResNet) for its training, thereby enabling the elimination of known adversarial perturbations. PPSA introduces model difference quantification and negative momentum to strategically schedule multiple pre-trained models, thereby maximizing the differences among adversarial examples during the defense model's training and enhancing its robustness in eliminating adversarial perturbations. MAPE effectively eliminates adversarial perturbations in various adversarial examples, providing a robust defense against attacks from different substitute models. In a black-box attack scenario utilizing ResNet-34 as the target model, our approach achieves average defense rates of over 95.1\% on CIFAR-10 and over 71.5\% on Mini-ImageNet, demonstrating state-of-the-art performance.
Continual learning (CL), where a model is trained on a sequence of data tasks, is increasingly being adopted across key fields such as large language models and image recognition, yet it remains highly vulnerable to data poisoning that triggers learning divergence or severe excess risk. Despite these threats, a principled theoretical foundation in CL for understanding attack and defense remains lacking. In this paper, we develop a theoretical framework to analyze strategic attacks and defenses in regularization-based CL, a cornerstone of recent CL theory. By framing the adversary-defender interaction as an online zero-sum game, we first establish a fundamental performance limit: no defense succeeds when an adversary poisons a linear proportion of tasks by injecting unbounded noise or pattern shifts in regularization-based CL. We then analyze two possibly defensible scenarios: infrequent attacks and bounded noise per attack. For the former regime, we propose a task-to-task verification mechanism to detect data poisoning and reduce cumulative bias for learning convergence. For the latter regime, we derive a robust defense that minimizes the model's sensitivity to poisoned features, provably accelerating the convergence rate. Extensive experiments on realistic tasks further validate our theoretical results.
Xingwei Zhong, Varun Sharma, Kar Wai Fok +1cs.CR cs.CV
Vision language models (VLMs) employ both visual and textual modalities to enable advanced vision-language inference. However, incorporating visual modalities expands the attack surface of VLMs, making them more susceptible to security threats such as adversarial perturbations and indirect prompt injection, wherein crafted malicious image prompts can elicit unintended model outputs. Existing defense methods against malicious image prompts remain insufficient as they typically demand extensive datasets for retraining or the deployment of additional, complex classifiers. Most critically, there is a profound lack of specialized defense mechanisms specifically targeting indirect prompt injections, a gap that serves as a primary motivation for this work. To address these limitations, we introduce DE-FIVE, a novel training-free framework for detecting malicious image prompts by leveraging Fourier features and the hidden state representations of the visual encoder (image vector embeddings) across perturbations. Specifically, we develop a hybrid detection strategy consisting of a black-box detector that operates on Fourier-domain features and a white-box detector that exploits image vector embeddings derived from only a few-shot malicious set. Extensive experiments demonstrate that the proposed framework consistently outperforms state-of-the-art baselines against malicious image prompts.
Farhin Farhad Riya, Shahinul Hoque, Yingyuan Yang +2cs.LG
Deep Neural Networks DNNs have achieved remarkable accuracy in various tasks including their application in CyberPhysical Systems CPS for detecting False Data Injection Attacks FDIA during critical operations However the unique infrastructure of CPS makes DNNs vulnerable to exploitation by attackers aiming to evade detection Additionally the distinct nature of CPS presents challenges for conventional defense mechanisms against FDIA This paper proposes an innovative defense framework that strengthens DNNs against such attacks by introducing an additional input layer that performs padding in the input samples using pseudofeature values derived from the inputs statistical distribution This padding increases the input dimensionality in a randomized and dataaware manner making adversarial attacks computationally infeasible due to the nontransferable nature of crafted perturbations and the unpredictability of the padded structure Our method is lightweight modelagnostic and requires no modifications to the core architecture making it highly deployable in realworld CPS settings We evaluated our framework on critical power grid applications such as state estimation using the IEEE 14bus 30bus 118bus and 300bus systems Experiments under adversarial settings demonstrate that our padding strategy significantly improves model robustness with negligible impact on performance and effectively mitigates attacks that would otherwise bypass conventional defenses
While Latent Diffusion Models (LDMs) have revolutionized visual synthesis, they are increasingly exploited for unauthorized mimicry of individuals. Existing defenses inject deceptive perturbations to steer the generated images toward irrelevant targets. However, this approach hinges on an ungrounded assumption: subtle perturbations can maintain their deceptive efficacy throughout an LDM's extensive generation process. In reality, the model's innate restoration mechanism will remove such perturbations and cause individual identities to re-emerge in the images generated. We propose VOID, a defense framework that overcomes this conundrum by manipulating an LDM's intrinsic stochasticity. VOID perturbs the diffusion pipeline in two novel ways: 1) amplifying the latent encoding errors to shatter an image's semantic structure, and 2) counteracting the target guidance signals to suppress the model's restoration capabilities. This results in a semantic corruption that thwarts any unauthorized mimicry. Notably, the security gain does not come at the price of visual utility, as VOID simultaneously manages to confine perturbations to human-imperceptible regions of protected images. Our comprehensive evaluation of 24 state-of-the-art defenses against 10 mimicry attacks on 5 datasets demonstrates VOID's unprecedented protection power: it increases the average Frechet Inception Distance (FID) from 113 to 365, a 223% improvement over the strongest defense to date.
Self-supervised data curation provides a pathway to scaling and improving the generalization capabilities of machine learning models. By leveraging self-supervised learning (SSL) for data curation, the demand for massive training datasets required by foundation models can be effectively met. SSL greatly alleviates the costs associated with annotation and manual dataset curation while minimizing the need for human oversight. However, the integrity of SSL-curated datasets must be rigorously checked, as reliance on anonymous and unvetted external sources can substantially increase the risk of data poisoning. In this paper, we propose a Poisoned Data Detector (PDD), an active defense mechanism designed to ensure the integrity of SSL-curated datasets prior to foundation model training. PDDs are designed using a combination of the pretrained ImageBind model and traditional classifiers, including Random Forest (RF), k-Nearest Neighbors (KNN), Naive Bayes (NB), and Support Vector Machines (SVM). We rigorously evaluated PDDs using 176,200 images from three diverse datasets and three different adversarial attacks encompassing both in-distribution and out-of-distribution scenarios. Notably, SVM-PDD achieves superior performance for both in-distribution (Set3-Set5) and out-of-distribution (TrueFace and 140K RealFace) datasets. Our design demonstrates strong scalability and enables the rapid integration of new adversarial attack detectors through an ensemble approach.
LLM-based ranking systems are vulnerable to Generative Engine Optimization (GEO) attacks, where adversaries inject semantic signals into product descriptions to artificially boost rankings. We propose SCI-Defense, a three-component defense framework combining Perplexity detection (PPL), Semantic Integrity Scoring (SIS), and Inter-Candidate Detection (ICD). SIS evaluates four manipulation dimensions: Authority Attribution (AA), Narrative Purposiveness (NP), Comparative Claims (CA), and Temporal Claims (TC). Evaluated on 600 Amazon product descriptions across 6 categories, SCI-Defense achieves Precision=1.000 and FPR=0.000, with Recall of 1.000, 0.952, and 0.830 against String, Reasoning, and Review attacks respectively. On 600 MS MARCO web passages, String attacks are blocked with perfect recall while Review attacks yield near-zero recall, as web passages lack the persuasion-oriented signals that SIS targets in product descriptions. We demonstrate that existing defenses -- PPL-only filters, SafetyClf content classifiers, and paraphrasing -- achieve zero recall against semantic manipulation attacks. We further demonstrate new attacks such as Specification Amplification and Use-Case Saturation can expose semantic relevance manipulation as a structural defense blind spot that suggests directions for future research.
Emma Andrews, Sahan Sanjaya, Prabhat Mishraquant-ph cs.LG
Machine learning models can learn from data samples to carry out various tasks efficiently. When data samples are adversarially manipulated, such as by insertion of carefully crafted noise, it can cause the model to make mistakes. Quantum machine learning models are also vulnerable to such adversarial attacks, especially in image classification using variational quantum classifiers. While there are promising defenses against these adversarial perturbations, such as training with adversarial samples, they face practical limitations. For example, they are not applicable in scenarios where training with adversarial samples is either not possible or can overfit the models on one type of attack. In this paper, we propose an adversarial training-free defense framework that utilizes a quantum autoencoder to purify the adversarial samples through reconstruction. Moreover, our defense framework provides a confidence metric to identify potentially adversarial samples that cannot be purified the quantum autoencoder. Extensive evaluation demonstrates that our defense framework can significantly outperform state-of-the-art in prediction accuracy (up to 68%) under adversarial attacks.