Advancements in wireless and mobile technologies, including 5G advanced and the envisioned 6G, are driving exponential growth in wireless devices. However, this rapid expansion exacerbates spectrum scarcity, posing a critical challenge. Dynamic spectrum allocation (DSA)--which relies on sensing and dynamically sharing spectrum--has emerged as an essential solution to address this issue. While machine learning (ML) models hold significant potential for improving spectrum sensing, their adoption in centralized ML-based DSA systems is limited by privacy concerns, bandwidth constraints, and regulatory challenges. To overcome these limitations, distributed ML-based approaches such as Federated Learning (FL) offer promising alternatives. This work addresses two key challenges in FL-based spectrum sensing (FLSS). First, the scarcity of labeled data for training FL models in practical spectrum sensing scenarios is tackled with a semi-supervised FL approach, combined with energy detection, enabling model training on unlabeled datasets. Second, we examine the security vulnerabilities of FLSS, focusing on the impact of data poisoning attacks. Our analysis highlights the shortcomings of existing majority-based defenses in countering such attacks. To address these vulnerabilities, we propose a novel defense mechanism inspired by vaccination, which effectively mitigates data poisoning attacks without relying on majority-based assumptions. Extensive experiments on both synthetic and real-world datasets validate our solutions, demonstrating that FLSS can achieve near-perfect accuracy on unlabeled datasets and maintain Byzantine robustness against both targeted and untargeted data poisoning attacks, even when a significant proportion of participants are malicious.
Federated learning (FL) enables collaborative model training across decentralized clients while preserving data privacy. However, its open-participation nature exposes it to data-poisoning attacks, in which malicious actors submit corrupted model updates to degrade the global model. Existing defenses are often reactive, relying on statistical aggregation rules that can be computationally expensive and that typically assume an honest majority. This paper introduces a proactive, economic defense: a lightweight Bayesian incentive mechanism that makes malicious behavior economically irrational. Each training round is modeled as a Bayesian game of incomplete information in which the server, acting as the principal, uses a small, private validation dataset to verify update quality before issuing payments. The design satisfies Individual Rationality (IR) for benevolent clients, ensuring their participation is profitable, and Incentive Compatibility (IC), making poisoning an economically dominated strategy. Extensive experiments on non-IID partitions of MNIST and FashionMNIST demonstrate robustness: with 50% label-flipping adversaries on MNIST, the mechanism maintains 96.7% accuracy, only 0.3 percentage points lower than in a scenario with 30% label-flipping adversaries. This outcome is 51.7 percentage points better than standard FedAvg, which collapses under the same 50% attack. The mechanism is computationally light, budget-bounded, and readily integrates into existing FL frameworks, offering a practical route to economically robust and sustainable FL ecosystems.
Recent studies have shown that Large Language Models (LLMs) are vulnerable to data poisoning attacks, where malicious training examples embed hidden behaviours triggered by specific input patterns. However, most existing works assume a phrase and focus on the attack's effectiveness, offering limited understanding of trigger mechanisms and how multiple triggers interact within the model. In this paper, we present a framework for studying poisoning in LLMs. We show that multiple distinct backdoor triggers can coexist within a single model without interfering with each other, enabling adversaries to embed several triggers concurrently. Using multiple triggers with high embedding similarity, we demonstrate that poisoned triggers can achieve robust activation even when tokens are substituted or separated by long token spans. Our findings expose a broader and more persistent vulnerability surface in LLMs. To mitigate this threat, we propose a post hoc recovery method that selectively retrains specific model components based on a layer-wise weight difference analysis. Our method effectively removes the trigger behaviour with minimal parameter updates, presenting a practical and efficient defence against multi-trigger poisoning.
Our research addresses the overlooked security concerns related to data poisoning in continual learning (CL). Data poisoning - the intentional manipulation of training data to affect the predictions of machine learning models - was recently shown to be a threat to CL training stability. While existing literature predominantly addresses scenario-dependent attacks, we propose to focus on a more simple and realistic single-task poison (STP) threats. In contrast to previously proposed poisoning settings, in STP adversaries lack knowledge and access to the model, as well as to both previous and future tasks. During an attack, they only have access to the current task within the data stream. Our study demonstrates that even within these stringent conditions, adversaries can compromise model performance using standard image corruptions. We show that STP attacks are able to strongly disrupt the whole continual training process: decreasing both the stability (its performance on past tasks) and plasticity (capacity to adapt to new tasks) of the algorithm. Finally, we propose a high-level defense framework for CL along with a poison task detection method based on task vectors. The code is available at https://github.com/stapaw/STP.git .
Gradient-based optimization is the workhorse of deep learning, offering efficient and scalable training via backpropagation. However, its reliance on large volumes of labeled data raises privacy and security concerns such as susceptibility to data poisoning attacks and the risk of overfitting. In contrast, black box optimization methods, which treat the model as an opaque function, relying solely on function evaluations to guide optimization, offer a promising alternative in scenarios where data access is restricted, adversarial risks are high, or overfitting is a concern. However, black box methods also pose significant challenges, including poor scalability to high-dimensional parameter spaces, as prevalent in large language models (LLMs), and high computational costs due to reliance on numerous model evaluations. This paper introduces BBoxER, an evolutionary black-box method for LLM post-training that induces an information bottleneck via implicit compression of the training data. Leveraging the tractability of information flow, we provide strong theoretical bounds on generalization, differential privacy, susceptibility to data poisoning attacks, and robustness to extraction attacks. BBoxER operates on top of pre-trained LLMs, offering a lightweight and modular enhancement suitable for deployment in restricted or privacy-sensitive environments, in addition to non-vacuous generalization guarantees. In experiments with LLMs, we demonstrate empirically that Retrofitting methods are able to learn, showing how a few iterations of BBoxER improve performance and generalize well on a benchmark of reasoning datasets. This positions BBoxER as an attractive add-on on top of gradient-based optimization.
The pre-training of large language models (LLMs) relies on massive text datasets sourced from diverse and difficult-to-curate origins. Although membership inference attacks and hidden canaries have been explored to trace data usage, such methods rely on memorization of training data, which LM providers try to limit. In this work, we demonstrate that indirect data poisoning (where the targeted behavior is absent from training data) is not only feasible but also allow to effectively protect a dataset and trace its use. Using gradient-based optimization prompt-tuning, we make a model learn arbitrary secret sequences: secret responses to secret prompts that are absent from the training corpus. We validate our approach on language models pre-trained from scratch and show that less than 0.005% of poisoned tokens are sufficient to covertly make a LM learn a secret and detect it with extremely high confidence ($p < 10^{-55}$) with a theoretically certifiable scheme. Crucially, this occurs without performance degradation (on LM benchmarks) and despite secrets never appearing in the training set.
Robust distributed learning algorithms aim to maintain good performance in distributed and federated settings, even in the presence of misbehaving workers. Two primary threat models have been studied: Byzantine attacks, where misbehaving workers can send arbitrarily corrupted updates, and data poisoning attacks, where misbehavior is limited to manipulation of local training data. While prior work has shown comparable optimization error under both threat models, a fundamental question remains open: How do these threat models impact generalization? Empirical evidence suggests a gap between the two threat models, yet it remains unclear whether it is fundamental or merely an artifact of suboptimal attacks. In this work, we present the first theoretical investigation into this problem, formally showing that Byzantine attacks are intrinsically more harmful to generalization than data poisoning. Specifically, we prove that: (i) under data poisoning, the uniform algorithmic stability of a robust distributed learning algorithm, with optimal optimization error, degrades by an additive factor of $\varTheta ( \frac{f}{n-f} )$, with $f$ the number of misbehaving workers out of $n$; and (ii) In contrast, under Byzantine attacks, the degradation is in $\mathcal{O} \big( \sqrt{ \frac{f}{n-2f}} \big)$.This difference in stability leads to a generalization error gap that is especially significant as $f$ approaches its maximum value $\frac{n}{2}$.
Graph neural networks (GNNs) have achieved significant success in graph representation learning and have been applied to various domains. However, many real-world graphs contain sensitive personal information, such as user profiles in social networks, raising serious privacy concerns when graph learning is performed using GNNs. To address this issue, locally private graph learning protocols have gained considerable attention. These protocols leverage the privacy advantages of local differential privacy (LDP) and the effectiveness of GNN's message-passing in calibrating noisy data, offering strict privacy guarantees for users' local data while maintaining high utility (e.g., node classification accuracy) for graph learning. Despite these advantages, such protocols may be vulnerable to data poisoning attacks, a threat that has not been considered in previous research. Identifying and addressing these threats is crucial for ensuring the robustness and security of privacy-preserving graph learning frameworks. This work introduces the first data poisoning attack targeting locally private graph learning protocols. The attacker injects fake users into the protocol, manipulates these fake users to establish links with genuine users, and sends carefully crafted data to the server, ultimately compromising the utility of private graph learning. The effectiveness of the attack is demonstrated both theoretically and empirically. In addition, several defense strategies have also been explored, but their limited effectiveness highlights the need for more robust defenses.
Deep Neural Networks (DNNs) are vulnerable to backdoor attacks, where attackers implant hidden triggers during training to maliciously control model behavior. Topological Evolution Dynamics (TED) has recently emerged as a powerful tool for detecting backdoor attacks in DNNs. However, TED can be vulnerable to backdoor attacks that adaptively distort topological representation distributions across network layers. To address this limitation, we propose TED-LaST (Topological Evolution Dynamics against Laundry, Slow release, and Target mapping attack strategies), a novel defense strategy that enhances TED's robustness against adaptive attacks. TED-LaST introduces two key innovations: label-supervised dynamics tracking and adaptive layer emphasis. These enhancements enable the identification of stealthy threats that evade traditional TED-based defenses, even in cases of inseparability in topological space and subtle topological perturbations. We review and classify data poisoning tricks in state-of-the-art adaptive attacks and propose enhanced adaptive attack with target mapping, which can dynamically shift malicious tasks and fully leverage the stealthiness that adaptive attacks possess. Our comprehensive experiments on multiple datasets (CIFAR-10, GTSRB, and ImageNet100) and model architectures (ResNet20, ResNet101) show that TED-LaST effectively counteracts sophisticated backdoors like Adap-Blend, Adapt-Patch, and the proposed enhanced adaptive attack. TED-LaST sets a new benchmark for robust backdoor detection, substantially enhancing DNN security against evolving threats.
Chain of Thought (CoT) has been applied to various large language models (LLMs) and proven to be effective in improving the quality of outputs. In recent studies, transformers are proven to have absolute upper bounds in terms of expressive power, and consequently, they cannot solve many computationally difficult problems. However, empowered by CoT, transformers are proven to be able to solve some difficult problems effectively, such as the $k$-parity problem. Nevertheless, those works rely on two imperative assumptions: (1) identical training and testing distribution, and (2) corruption-free training data with correct reasoning steps. However, in the real world, these assumptions do not always hold. Although the risks of data shifts have caught attention, our work is the first to rigorously study the exact harm caused by such shifts to the best of our knowledge. Focusing on the $k$-parity problem, in this work we investigate the joint impact of two types of data shifts: the distribution shifts and data poisoning, on the quality of trained models obtained by a well-established CoT decomposition. In addition to revealing a surprising phenomenon that CoT leads to worse performance on learning parity than directly generating the prediction, our technical results also give a rigorous and comprehensive explanation of the mechanistic reasons of such impact.
Vision Language Models (VLMs) have shown remarkable performance, but are also vulnerable to backdoor attacks whereby the adversary can manipulate the model's outputs through hidden triggers. Prior attacks primarily rely on single-modality triggers, leaving the crucial cross-modal fusion nature of VLMs largely unexplored. Unlike prior work, we identify a novel attack surface that leverages cross-modal semantic mismatches as implicit triggers. Based on this insight, we propose BadSem (Backdoor Attack with Semantic Manipulation), a data poisoning attack that injects stealthy backdoors by deliberately misaligning image-text pairs during training. To perform the attack, we construct SIMBad, a dataset tailored for semantic manipulation involving color and object attributes. Extensive experiments across four widely used VLMs show that BadSem achieves over 98% average ASR, generalizes well to out-of-distribution datasets, and can transfer across poisoning modalities. Our detailed analysis using attention visualization shows that backdoored models focus on semantically sensitive regions under mismatched conditions while maintaining normal behavior on clean inputs. To mitigate the attack, we try two defense strategies based on system prompt and supervised fine-tuning but find that both of them fail to mitigate the semantic backdoor. Our findings highlight the urgent need to address semantic vulnerabilities in VLMs for their safer deployment.
Deep Neural Networks (DNNs) are widely used for traffic sign recognition because they can automatically extract high-level features from images. These DNNs are trained on large-scale datasets obtained from unknown sources. Therefore, it is important to ensure that the models remain secure and are not compromised or poisoned during training. In this paper, we investigate the robustness of DNNs trained for traffic sign recognition. First, we perform the error-minimizing attacks on DNNs used for traffic sign recognition by adding imperceptible perturbations on training data. Then, we propose a data augmentation-based training method to mitigate the error-minimizing attacks. The proposed training method utilizes nonlinear transformations to disrupt the perturbations and improve the model robustness. We experiment with two well-known traffic sign datasets to demonstrate the severity of the attack and the effectiveness of our mitigation scheme. The error-minimizing attacks reduce the prediction accuracy of the DNNs from 99.90% to 10.6%. However, our mitigation scheme successfully restores the prediction accuracy to 96.05%. Moreover, our approach outperforms adversarial training in mitigating the error-minimizing attacks. Furthermore, we propose a detection model capable of identifying poisoned data even when the perturbations are imperceptible to human inspection. Our detection model achieves a success rate of over 99% in identifying the attack. This research highlights the need to employ advanced training methods for DNNs in traffic sign recognition systems to mitigate the effects of data poisoning attacks.
One way to mitigate risks in vision-language models (VLMs) is to remove dangerous samples in their training data. However, such data moderation can be easily bypassed when harmful images are split into small, benign-looking patches, scattered across many training samples. VLMs may then learn to piece these fragments together during training and generate harmful responses at inference, either from full images or text references. For instance, if trained on image patches from a bloody scene paired with the descriptions "safe," VLMs may later describe, the full image or a text reference to the scene, as "safe." We define the core ability of VLMs enabling this attack as $\textit{visual stitching}$ -- the ability to integrate visual information spread across multiple training samples that share the same textual descriptions. In our work, we first demonstrate visual stitching abilities in common open-source VLMs on three datasets where each image is labeled with a unique synthetic ID: we split each $(\texttt{image}, \texttt{ID})$ pair into $\{(\texttt{patch}, \texttt{ID})\}$ pairs at different granularity for finetuning, and we find that tuned models can verbalize the correct IDs from full images or text reference. Building on this, we simulate the adversarial data poisoning scenario mentioned above by using patches from dangerous images and replacing IDs with text descriptions like ``safe'' or ``unsafe'', demonstrating how harmful content can evade moderation in patches and later be reconstructed through visual stitching, posing serious VLM safety risks. Code is available at https://github.com/ZHZisZZ/visual-stitching.
Retrieval-Augmented Generation (RAG) systems, which integrate Large Language Models (LLMs) with external knowledge sources, are vulnerable to a range of adversarial attack vectors. This paper examines the importance of RAG systems through recent industry adoption trends and identifies the prominent attack vectors for RAG: prompt injection, data poisoning, and adversarial query manipulation. We analyze these threats under risk management lens, and propose robust prioritized control list that includes risk-mitigating actions like input validation, adversarial training, and real-time monitoring.
Model distillation has become essential for creating smaller, deployable language models that retain larger system capabilities. However, widespread deployment raises concerns about resilience to adversarial manipulation. This paper investigates vulnerability of distilled models to adversarial injection of biased content during training. We demonstrate that adversaries can inject subtle biases into teacher models through minimal data poisoning, which propagates to student models and becomes significantly amplified. We propose two propagation modes: Untargeted Propagation, where bias affects multiple tasks, and Targeted Propagation, focusing on specific tasks while maintaining normal behavior elsewhere. With only 25 poisoned samples (0.25% poisoning rate), student models generate biased responses 76.9% of the time in targeted scenarios - higher than 69.4% in teacher models. For untargeted propagation, adversarial bias appears 6x-29x more frequently in student models on unseen tasks. We validate findings across six bias types (targeted advertisements, phishing links, narrative manipulations, insecure coding practices), various distillation methods, and different modalities spanning text and code generation. Our evaluation reveals shortcomings in current defenses - perplexity filtering, bias detection systems, and LLM-based autorater frameworks - against these attacks. Results expose significant security vulnerabilities in distilled models, highlighting need for specialized safeguards. We propose practical design principles for building effective adversarial bias mitigation strategies.
In connected and autonomous vehicles, machine learning for safety message classification has become critical for detecting malicious or anomalous behavior. However, conventional approaches that rely on centralized data collection or purely local training face limitations due to the large scale, high mobility, and heterogeneous data distributions inherent in inter-vehicle networks. To overcome these challenges, this paper explores Distributed Federated Learning (DFL), whereby vehicles collaboratively train deep learning models by exchanging model updates among one-hop neighbors and propagating models over multiple hops. Using the Vehicular Reference Misbehavior (VeReMi) Extension Dataset, we show that DFL can significantly improve classification accuracy across all vehicles compared to learning strictly with local data. Notably, vehicles with low individual accuracy see substantial accuracy gains through DFL, illustrating the benefit of knowledge sharing across the network. We further show that local training data size and time-varying network connectivity correlate strongly with the model's overall accuracy. We investigate DFL's resilience and vulnerabilities under attacks in multiple domains, namely wireless jamming and training data poisoning attacks. Our results reveal important insights into the vulnerabilities of DFL when confronted with multi-domain attacks, underlining the need for more robust strategies to secure DFL in vehicular networks.
Time Series Foundation Models (TSFMs), which are pretrained on large-scale, cross-domain data and capable of zero-shot forecasting in new scenarios without further training, are increasingly adopted in real-world applications. However, as the zero-shot forecasting paradigm gets popular, a critical yet overlooked question emerges: Are TSFMs robust to adversarial input perturbations? Such perturbations could be exploited in man-in-the-middle attacks or data poisoning. To address this gap, we conduct a systematic investigation into the adversarial robustness of TSFMs. Our results show that even minimal perturbations can induce significant and controllable changes in forecast behaviors, including trend reversal, temporal drift, and amplitude shift, posing serious risks to TSFM-based services. Through experiments on representative TSFMs and multiple datasets, we reveal their consistent vulnerabilities and identify potential architectural designs, such as structural sparsity and multi-task pretraining, that may improve robustness. Our findings offer actionable guidance for designing more resilient forecasting systems and provide a critical assessment of the adversarial robustness of TSFMs.
Large Language Models (LLMs) such as GPT-4 and its recent iterations, Google's Gemini, Anthropic's Claude 3 models, and xAI's Grok have caused a revolution in natural language processing, but their capabilities also introduce new security vulnerabilities. In this survey, we provide a comprehensive overview of the emerging security concerns around LLMs, categorizing threats into prompt injection and jailbreaking, adversarial attacks such as input perturbations and data poisoning, misuse by malicious actors for purposes such as generating disinformation, phishing emails, and malware, and worrisome risks inherent in autonomous LLM agents. A significant focus has been recently placed on the latter, exploring goal misalignment, emergent deception, self-preservation instincts, and the potential for LLMs to develop and pursue covert, misaligned objectives, a behavior known as scheming, which may even persist through safety training. We summarize recent academic and industrial studies from 2022 to 2025 that exemplify each threat, analyze proposed defenses and their limitations, and identify open challenges in securing LLM-based applications. We conclude by emphasizing the importance of advancing robust, multi-layered security strategies to ensure LLMs are safe and beneficial.
Backdoor attacks, or trojans, pose a security risk by concealing undesirable behavior in deep neural network models. Open-source neural networks are downloaded from the internet daily, possibly containing backdoors, and third-party model developers are common. To advance research on backdoor attack mitigation, we develop several trojans for deep reinforcement learning (DRL) agents. We focus on in-distribution triggers, which occur within the agent's natural data distribution, since they pose a more significant security threat than out-of-distribution triggers due to their ease of activation by the attacker during model deployment. We implement backdoor attacks in four reinforcement learning (RL) environments: LavaWorld, Randomized LavaWorld, Colorful Memory, and Modified Safety Gymnasium. We train various models, both clean and backdoored, to characterize these attacks. We find that in-distribution triggers can require additional effort to implement and be more challenging for models to learn, but are nevertheless viable threats in DRL even using basic data poisoning attacks.
We investigate the theoretical foundations of data poisoning attacks in machine learning models. Our analysis reveals that the Hessian with respect to the input serves as a diagnostic tool for detecting poisoning, exhibiting spectral signatures that characterize compromised datasets. We use random matrix theory (RMT) to develop a theory for the impact of poisoning proportion and regularisation on attack efficacy in linear regression. Through QR stepwise regression, we study the spectral signatures of the Hessian in multi-output regression. We perform experiments on deep networks to show experimentally that this theory extends to modern convolutional and transformer networks under the cross-entropy loss. Based on these insights we develop preliminary algorithms to determine if a network has been poisoned and remedies which do not require further training.