DNNs are susceptible to defects like backdoors, adversarial attacks, and unfairness, undermining their reliability. Existing approaches mainly involve retraining, optimization, constraint-solving, or search algorithms. However, most methods rely on gradient calculations, restricting applicability to specific activation functions (e.g., ReLU), or use search algorithms with uninterpretable localization and repair. Furthermore, they often lack generalizability across multiple properties. We propose SHARPEN, integrating interpretable fault localization with a derivative-free optimization strategy. First, SHARPEN introduces a Deep SHAP-based localization strategy quantifying each layer's and neuron's marginal contribution to erroneous outputs. Specifically, a hierarchical coarse-to-fine approach reranks layers by aggregated impact, then locates faulty neurons/filters by analyzing activation divergences between property-violating and benign states. Subsequently, SHARPEN incorporates CMA-ES to repair identified neurons. CMA-ES leverages a covariance matrix to capture variable dependencies, enabling gradient-free search and coordinated adjustments across coupled neurons. By combining interpretable localization with evolutionary optimization, SHARPEN enables derivative-free repair across architectures, being less sensitive to gradient anomalies and hyperparameters. We demonstrate SHARPEN's effectiveness on three repair tasks. Balancing property repair and accuracy preservation, it outperforms baselines in backdoor removal (+10.56%), adversarial mitigation (+5.78%), and unfairness repair (+11.82%). Notably, SHARPEN handles diverse tasks, and its modular design is plug-and-play with different derivative-free optimizers, highlighting its flexibility.

World models -- learned internal simulators of environment dynamics -- are rapidly becoming foundational to autonomous decision-making in robotics, autonomous vehicles, and agentic AI. Yet this predictive power introduces a distinctive set of safety, security, and cognitive risks. Adversaries can corrupt training data, poison latent representations, and exploit compounding rollout errors to cause catastrophic failures in safety-critical deployments. World model-equipped agents are more capable of goal misgeneralisation, deceptive alignment, and reward hacking precisely because they can simulate the consequences of their own actions. Authoritative world model predictions further foster automation bias and miscalibrated human trust that operators lack the tools to audit. This paper surveys the world model landscape; introduces formal definitions of trajectory persistence and representational risk; presents a five-profile attacker capability taxonomy; and develops a unified threat model extending MITRE ATLAS and the OWASP LLM Top 10 to the world model stack. We provide an empirical proof-of-concept on trajectory-persistent adversarial attacks (GRU-RSSM: A_1 = 2.26x amplification, -59.5% reduction under adversarial fine-tuning; stochastic RSSM proxy: A_1 = 0.65x; DreamerV3 checkpoint: non-zero action drift confirmed). We illustrate risks through four deployment scenarios and propose interdisciplinary mitigations spanning adversarial hardening, alignment engineering, NIST AI RMF and EU AI Act governance, and human-factors design. We argue that world models must be treated as safety-critical infrastructure requiring the same rigour as flight-control software or medical devices.

Model merging has emerged as a powerful technique for combining specialized capabilities from multiple fine-tuned LLMs without additional training costs. However, the security implications of this widely-adopted practice remain critically underexplored. In this work, we reveal that model merging introduces a novel attack surface that can be systematically exploited to compromise safety alignment. We present TrojanMerge,, a framework that embeds latent malicious components into source models that remain individually benign but produce severely misaligned models when merged. Our key insight is formulating this attack as a constrained optimization problem: we construct perturbations that preserve source model safety through directional consistency constraints, maintain capabilities via Frobenius directional alignment constraints, yet combine during merging to form pre-computed attack vectors. Extensive experiments across 9 LLMs from 3 model families demonstrate that TrojanMerge, consistently achieves high harmful response rates in merged models while source models maintain safety scores comparable to unmodified versions. Our attack succeeds across diverse merging algorithms and remains effective under various hyperparameter configurations. These findings expose fundamental vulnerabilities in current model merging practices and highlight the urgent need for security-aware mechanisms.

A new generation of language models reasons entirely in continuous hidden states, producing no tokens and leaving no audit trail. We show that this silence creates a fundamentally new attack surface. ThoughtSteer perturbs a single embedding vector at the input layer; the model's own multi-pass reasoning amplifies this perturbation into a hijacked latent trajectory that reliably produces the attacker's chosen answer, while remaining structurally invisible to every token-level defense. Across two architectures (Coconut and SimCoT), three reasoning benchmarks, and model scales from 124M to 3B parameters, ThoughtSteer achieves >=99% attack success rate with near-baseline clean accuracy, transfers to held-out benchmarks without retraining (94-100%), evades all five evaluated active defenses, and survives 25 epochs of clean fine-tuning. We trace these results to a unifying mechanism: Neural Collapse in the latent space pulls triggered representations onto a tight geometric attractor, explaining both why defenses fail and why any effective backdoor must leave a linearly separable signature (probe AUC>=0.999). Yet a striking paradox emerges: individual latent vectors still encode the correct answer even as the model outputs the wrong one. The adversarial information is not in any single vector but in the collective trajectory, establishing backdoor perturbations as a new lens for mechanistic interpretability of continuous reasoning. Code and checkpoints are available.

Backdoor attacks on federated learning (FL) are most often evaluated with synthetic corner patches or out-of-distribution (OOD) patterns that are unlikely to arise in practice. In this paper, we revisit the backdoor threat to standard FL (a single global model) under a more realistic setting where triggers must be semantically meaningful, in-distribution, and visually plausible. We propose SABLE, a Semantics-Aware Backdoor for LEarning in federated settings, which constructs natural, content-consistent triggers (e.g., semantic attribute changes such as sunglasses) and optimizes an aggregation-aware malicious objective with feature separation and parameter regularization to keep attacker updates close to benign ones. We instantiate SABLE on CelebA hair-color classification and the German Traffic Sign Recognition Benchmark (GTSRB), poisoning only a small, interpretable subset of each malicious client's local data while otherwise following the standard FL protocol. Across heterogeneous client partitions and multiple aggregation rules (FedAvg, Trimmed Mean, MultiKrum, and FLAME), our semantics-driven triggers achieve high targeted attack success rates while preserving benign test accuracy. These results show that semantics-aligned backdoors remain a potent and practical threat in federated learning, and that robustness claims based solely on synthetic patch triggers can be overly optimistic.

Dataset Condensation (DC) is a data-efficient learning paradigm that synthesizes small yet informative datasets, enabling models to match the performance of full-data training. However, recent work exposes a critical vulnerability of DC to backdoor attacks, where malicious patterns (\textit{e.g.}, triggers) are implanted into the condensation dataset, inducing targeted misclassification on specific inputs. Existing attacks always prioritize attack effectiveness and model utility, overlooking the crucial dimension of stealthiness. To bridge this gap, we propose InkDrop, which enhances the imperceptibility of malicious manipulation without degrading attack effectiveness and model utility. InkDrop leverages the inherent uncertainty near model decision boundaries, where minor input perturbations can induce semantic shifts, to construct a stealthy and effective backdoor attack. Specifically, InkDrop first selects candidate samples near the target decision boundary that exhibit latent semantic affinity to the target class. It then learns instance-dependent perturbations constrained by perceptual and spatial consistency, embedding targeted malicious behavior into the condensed dataset. Extensive experiments across diverse datasets validate the overall effectiveness of InkDrop, demonstrating its ability to integrate adversarial intent into condensed datasets while preserving model utility and minimizing detectability. Our code is available at https://github.com/lvdongyi/InkDrop.

Federated Learning (FL) is witnessing wider adoption due to its ability to benefit from large amounts of scattered data while preserving privacy. However, despite its advantages, federated learning suffers from several setbacks that directly impact the accuracy, and the integrity of the global model it produces. One of these setbacks is the presence of malicious clients who actively try to harm the global model by injecting backdoor data into their local models while trying to evade detection. The objective of such clients is to trick the global model into making false predictions during inference, thereby compromising the integrity and trustworthiness of the global model on which honest stakeholders rely. To mitigate such mischievous behavior, we propose FedBBA (Federated Backdoor and Behavior Analysis). The proposed model aims to dampen the effect of such clients on the final accuracy, creating more resilient federated learning environments. We engineer our approach through the combination of (1) a reputation system to evaluate and track client behavior, (2) an incentive mechanism to reward honest participation and penalize malicious behavior, and (3) game theoretical models with projection pursuit analysis (PPA) to dynamically identify and minimize the impact of malicious clients on the global model. Extensive simulations on the German Traffic Sign Recognition Benchmark (GTSRB) and Belgium Traffic Sign Classification (BTSC) datasets demonstrate that FedBBA reduces the backdoor attack success rate to approximately 1.1%--11% across various attack scenarios, significantly outperforming state-of-the-art defenses like RDFL and RoPE, which yielded attack success rates between 23% and 76%, while maintaining high normal task accuracy (~95%--98%).

Backdoor attacks pose a significant threat to the integrity and reliability of Artificial Intelligence (AI) models, enabling adversaries to manipulate model behavior by injecting poisoned data with hidden triggers. These attacks can lead to severe consequences, especially in critical applications such as autonomous driving, healthcare, and finance. Detecting and mitigating backdoor attacks is crucial across the lifespan of model's phases, including pre-training, in-training, and post-training. In this paper, we propose Pre-Training Backdoor Mitigation for Federated Learning (FL-PBM), a novel defense mechanism that proactively filters poisoned data on the client side before model training in a federated learning (FL) environment. The approach consists of three stages: (1) inserting a benign trigger into the data to establish a controlled baseline, (2) applying Principal Component Analysis (PCA) to extract discriminative features and assess the separability of the data, (3) performing Gaussian Mixture Model (GMM) clustering to identify potentially malicious data samples based on their distribution in the PCA-transformed space, and (4) applying a targeted blurring technique to disrupt potential backdoor triggers. Together, these steps ensure that suspicious data is detected early and sanitized effectively, thereby minimizing the influence of backdoor triggers on the global model. Experimental evaluations on image-based datasets demonstrate that FL-PBM reduces attack success rates by up to 95% compared to baseline federated learning (FedAvg) and by 30 to 80% relative to state-of-the-art defenses (RDFL and LPSF). At the same time, it maintains over 90% clean model accuracy in most experiments, achieving better mitigation without degrading model performance.

Vision-Language Models (VLMs) are increasingly deployed in consumer applications where users seek recommendations about products, dining, and services. We introduce Hidden Ads, a new class of backdoor attacks that exploit this recommendation-seeking behavior to inject unauthorized advertisements. Unlike traditional pattern-triggered backdoors that rely on artificial triggers such as pixel patches or special tokens, Hidden Ads activates on natural user behaviors: when users upload images containing semantic content of interest (e.g., food, cars, animals) and ask recommendation-seeking questions, the backdoored model provides correct, helpful answers while seamlessly appending attacker-specified promotional slogans. This design preserves model utility and produces natural-sounding injections, making the attack practical for real-world deployment in consumer-facing recommendation services. We propose a multi-tier threat framework to systematically evaluate Hidden Ads across three adversary capability levels: hard prompt injection, soft prompt optimization, and supervised fine-tuning. Our poisoned data generation pipeline uses teacher VLM-generated chain-of-thought reasoning to create natural trigger--slogan associations across multiple semantic domains. Experiments on three VLM architectures demonstrate that Hidden Ads achieves high injection efficacy with near-zero false positives while maintaining task accuracy. Ablation studies confirm that the attack is data-efficient, transfers effectively to unseen datasets, and scales to multiple concurrent domain-slogan pairs. We evaluate defenses including instruction-based filtering and clean fine-tuning, finding that both fail to remove the backdoor without causing significant utility degradation.

Deep learning (DL) has been widely studied for assisting applications of modern wireless communications. One of the applications is automatic modulation classification (AMC). However, DL models are found to be vulnerable to adversarial machine learning (AML) threats. One of the most persistent and stealthy threats is the backdoor (Trojan) attack. Nevertheless, most studied threats originate from other AI domains, such as computer vision (CV). Therefore, in this paper, a physical backdoor attack targeting the wireless signal before transmission is studied. The adversary is considered to be using explainable AI (XAI) to guide the placement of the trigger in the most vulnerable parts of the signal. Then, a class prototype combined with principal components is used to generate the trigger. The studied threat was found to be efficient in breaching multiple DL-based AMC models. The attack achieves high success rates for a wide range of SNR values and a small poisoning ratio.

Deep Learning (DL) has become a key technology that assists radio frequency (RF) signal classification applications, such as modulation classification. However, the DL models are vulnerable to adversarial machine learning threats, such as data manipulation attacks. We study a physical backdoor (Trojan) attack that targets a DL-based modulation classifier. In contrast to digital backdoor attacks, where digital triggers are injected into the training dataset, we use power amplifier (PA) non-linear distortions to create physical triggers before the dataset is formed. During training, the adversary manipulates amplitudes of RF signals and changes their labels to a target modulation scheme, training a backdoored model. At inference, the adversary aims to keep the backdoor attack inactive such that the backdoored model maintains high accuracy on test signals. However, if they apply the same manipulation used during training on these test signals, the backdoor is activated, and the model misclassifies these signals. We demonstrate that our proposed attack achieves high attack success rates with few manipulated RD signals for different noise levels. Furthermore, we test the resilience of the proposed attack to multiple defense techniques, and the results show that these techniques fail to mitigate the attack.

As machine learning (ML) systems expand in both scale and functionality, the security landscape has become increasingly complex, with a proliferation of attacks and defenses. However, existing studies largely treat these threats in isolation, lacking a coherent framework to expose their shared principles and interdependencies. This fragmented view hinders systematic understanding and limits the design of comprehensive defenses. Crucially, the two foundational assets of ML -- \textbf{data} and \textbf{models} -- are no longer independent; vulnerabilities in one directly compromise the other. The absence of a holistic framework leaves open questions about how these bidirectional risks propagate across the ML pipeline. To address this critical gap, we propose a \emph{unified closed-loop threat taxonomy} that explicitly frames model-data interactions along four directional axes. Our framework offers a principled lens for analyzing and defending foundation models. The resulting four classes of security threats represent distinct but interrelated categories of attacks: (1) Data$\rightarrow$Data (D$\rightarrow$D): including \emph{data decryption attacks and watermark removal attacks}; (2) Data$\rightarrow$Model (D$\rightarrow$M): including \emph{poisoning, harmful fine-tuning attacks, and jailbreak attacks}; (3) Model$\rightarrow$Data (M$\rightarrow$D): including \emph{model inversion, membership inference attacks, and training data extraction attacks}; (4) Model$\rightarrow$Model (M$\rightarrow$M): including \emph{model extraction attacks}. Our unified framework elucidates the underlying connections among these security threats and establishes a foundation for developing scalable, transferable, and cross-modal security strategies, particularly within the landscape of foundation models.

Mechanistic interpretability reveals that safety-critical behaviors (e.g., alignment, jailbreak, backdoor) in Large Language Models (LLMs) are grounded in specialized functional components. However, existing safety attribution methods struggle with generalization and reliability due to their reliance on heuristic, domain-specific metrics and search algorithms. To address this, we propose \ourmethod, a unified safety interpretability framework that identifies functionally complete safety circuits in LLMs via optimization. Unlike methods focusing on isolated heads or neurons, \ourmethod introduces differentiable binary masks to extract multi-granular circuits through gradient descent on safety datasets, while integrates Safety Circuit Tuning to utilize these sparse circuits for efficient safety fine-tuning. We validate \ourmethod in two key scenarios in LLM safety: \textbf{(1) backdoor attacks}, identifying a backdoor circuit with 0.42\% sparsity, whose ablation eradicates the Attack Success Rate (ASR) from 100\% $\to$ 0.4\% while retaining over 99\% general utility; \textbf{(2) safety alignment}, localizing an alignment circuit with 3.03\% heads and 0.79\% neurons, whose removal spikes ASR from 0.8\% $\to$ 96.9\%, whereas excluding this circuit during helpfulness fine-tuning maintains 96.5\% safety retention.

The rapid adoption of mobile graphical user interface (GUI) agents, which autonomously control applications and operating systems (OS), exposes new system-level attack surfaces. Existing backdoors against web GUI agents and general GenAI models rely on environmental injection or deceptive pop-ups to mislead the agent operation. However, these techniques do not work on screenshots-based mobile GUI agents due to the challenges of restricted trigger design spaces, OS background interference, and conflicts in multiple trigger-action mappings. We propose AgentRAE, a novel backdoor attack capable of inducing Remote Action Execution in mobile GUI agents using visually natural triggers (e.g., benign app icons in notifications). To address the underfitting caused by natural triggers and achieve accurate multi-target action redirection, we design a novel two-stage pipeline that first enhances the agent's sensitivity to subtle iconographic differences via contrastive learning, and then associates each trigger with a specific mobile GUI agent action through a backdoor post-training. Our extensive evaluation reveals that the proposed backdoor preserves clean performance with an attack success rate of over 90% across ten mobile operations. Furthermore, it is hard to visibly detect the benign-looking triggers and circumvents eight representative state-of-the-art defenses. These results expose an overlooked backdoor vector in mobile GUI agents, underscoring the need for defenses that scrutinize notification-conditioned behaviors and internal agent representations.

Federated Learning (FL), as a popular distributed learning paradigm, has shown outstanding performance in improving computational efficiency and protecting data privacy, and is widely applied in industrial image classification. However, due to its distributed nature, FL is vulnerable to threats from malicious clients, with poisoning attacks being a common threat. A major limitation of existing poisoning attack methods is their difficulty in bypassing model performance tests and defense mechanisms based on model anomaly detection. This often results in the detection and removal of poisoned models, which undermines their practical utility. To ensure both the performance of industrial image classification and attacks, we propose a targeted poisoning attack, PoiCGAN, based on feature-label collaborative perturbation. Our method modifies the inputs of the discriminator and generator in the Conditional Generative Adversarial Network (CGAN) to influence the training process, generating an ideal poison generator. This generator not only produces specific poisoned samples but also automatically performs label flipping. Experiments across various datasets show that our method achieves an attack success rate 83.97% higher than baseline methods, with a less than 8.87% reduction in the main task's accuracy. Moreover, the poisoned samples and malicious models exhibit high stealthiness.

Many learning systems now use graph data in which each node also contains text, such as papers with abstracts or users with posts. Because these texts often come from open platforms, an attacker may be able to quietly poison a small part of the training data and later make the model produce wrong predictions on demand. This paper studies that risk in a realistic setting where the attacker edits only node text and does not change the graph structure. We propose TAGBD, a text-only backdoor attack for text-attributed graphs. TAGBD first finds training nodes that are easier to influence, then generates natural-looking trigger text with the help of a shadow graph model, and finally injects the trigger by either replacing the original text or appending a short phrase. Experiments on three benchmark datasets show that the attack is highly effective, transfers across different graph models, and remains strong under common defenses. These results demonstrate that text alone is a practical attack channel in graph learning systems and suggest that future defenses should inspect both graph links and node content.

Forecasting plays a crucial role in modern safety-critical applications, such as space operations. However, the increasing use of deep forecasting models introduces a new security risk of trojan horse attacks, carried out by hiding a backdoor in the training data or directly in the model weights. Once implanted, the backdoor is activated by a specific trigger pattern at test time, causing the model to produce manipulated predictions. We focus on this issue in our \textit{Trojan Horse Hunt} data science competition, where more than 200 teams faced the task of identifying triggers hidden in deep forecasting models for spacecraft telemetry. We describe the novel task formulation, benchmark set, evaluation protocol, and best solutions from the competition. We further summarize key insights and research directions for effective identification of triggers in time series forecasting models. All materials are publicly available on the official competition webpage https://www.kaggle.com/competitions/trojan-horse-hunt-in-space.

Decentralized Federated Learning (DFL) remains highly vulnerable to adaptive backdoor attacks designed to bypass traditional passive defense metrics. To address this limitation, we shift the defensive paradigm toward a novel active, interventional auditing framework. First, we establish a dynamical model to characterize the spatiotemporal diffusion of adversarial updates across complex graph topologies. Second, we introduce a suite of proactive auditing metrics, stochastic entropy anomaly, randomized smoothing Kullback-Leibler divergence, and activation kurtosis. These metrics utilize private probes to stress-test local models, effectively exposing latent backdoors that remain invisible to conventional static detection. Furthermore, we implement a topology-aware defense placement strategy to maximize global aggregation resilience. We provide theoretical property for the system's convergence under co-evolving attack and defense dynamics. Numeric empirical evaluations across diverse architectures demonstrate that our active framework is highly competitive with state-of-the-art defenses in mitigating stealthy, adaptive backdoors while preserving primary task utility.

With the widespread deployment of deep-learning-based speech models in security-critical applications, backdoor attacks have emerged as a serious threat: an adversary who poisons a small fraction of training data can implant a hidden trigger that controls the model's output while preserving normal behavior on clean inputs. Existing inference-time defenses are not well suited to the audio domain, as they either rely on trigger over-robustness assumptions that fail on transformation-based and semantic triggers, or depend on properties specific to image or text modalities. In this paper, we propose STEP (Stability-based Trigger Exposure Profiling), a black-box, retraining-free backdoor detector that operates under hard-label-only access. Its core idea is to exploit a characteristic dual anomaly of backdoor triggers: anomalous label stability under semantic-breaking perturbations, and anomalous label fragility under semantic-preserving perturbations. STEP profiles each test sample with two complementary perturbation branches that target these two properties respectively, scores the resulting stability features with one-class anomaly detectors trained on benign references, and fuses the two scores via unsupervised weighting. Extensive experiments across seven backdoor attacks show that STEP achieves an average AUROC of 97.92% and EER of 4.54%, substantially outperforming state-of-the-art baselines, and generalizes across model architectures, speech tasks, an open-set verification scenario, and over-the-air physical-world settings.

Semantic segmentation models are widely deployed in safety-critical applications such as autonomous driving, yet their vulnerability to backdoor attacks remains largely underexplored. Prior segmentation backdoor studies transfer threat settings from existing image classification tasks, focusing primarily on object-to-background mis-segmentation. In this work, we revisit the threats by systematically examining backdoor attacks tailored to semantic segmentation. We identify four coarse-grained attack vectors (Object-to-Object, Object-to-Background, Background-to-Object, and Background-to-Background attacks), as well as two fine-grained vectors (Instance-Level and Conditional attacks). To formalize these attacks, we introduce BADSEG, a unified framework that optimizes trigger designs and applies label manipulation strategies to maximize attack performance while preserving victim model utility. Extensive experiments across diverse segmentation architectures on benchmark datasets demonstrate that BADSEG achieves high attack effectiveness with minimal impact on clean samples. We further evaluate six representative defenses and find that they fail to reliably mitigate our attacks, revealing critical gaps in current defenses. Finally, we demonstrate that these vulnerabilities persist in recent emerging architectures, including transformer-based networks and the Segment Anything Model (SAM), thereby compromising their security. Our work reveals previously overlooked security vulnerabilities in semantic segmentation, and motivates the development of defenses tailored to segmentation-specific threat models.