Adversarial Robustness in Machine Learning

Últimos artículos

• Probe-Based Data Attribution: Discovering and Mitigating Undesirable Behaviors in LLM Post-Training

  Frank Xiao, Santiago Aranguri · 28 de abril de 2026

  We propose probe-based data attribution, a method that traces behavioral changes in post-trained language models to responsible training datapoints. By computing activation-difference vectors for both test prompts and preference pairs and ranking by cosine similarity, we identify datapoints that cau…

• Unveiling the Backdoor Mechanism Hidden Behind Catastrophic Overfitting in Fast Adversarial Training

  Mengnan Zhao, Lihe Zhang, Tianhang Zheng, Bo Wang, Baocai Yin · 28 de abril de 2026

  Fast Adversarial Training (FAT) has attracted significant attention due to its efficiency in enhancing neural network robustness against adversarial attacks. However, FAT is prone to catastrophic overfitting (CO), wherein models overfit to the specific attack used during training and fail to general…

• Beyond Context: Large Language Models' Failure to Grasp Users' Intent

  Ahmed M. Hussain, Salahuddin Salahuddin · 28 de abril de 2026

  Current Large Language Models (LLMs) safety approaches focus on explicitly harmful content while overlooking a critical vulnerability: the inability to understand context and recognize user intent. This creates exploitable vulnerabilities that malicious users can systematically leverage to circumven…

• CAPSULE: Control-Theoretic Action Perturbations for Safe Uncertainty-Aware Reinforcement Learning

  Rahul Narava, Siddharth Verma, Ojas Jain, Shashi Shekhar Jha, Mayank Shekhar Jha · 28 de abril de 2026

  Ensuring safe exploration in high-dimensional systems with unknown dynamics remains a significant challenge. Existing safe reinforcement learning methods often provide safety guarantees only in expectation, which can still lead to safety violations. Control-theoretic approaches, in contrast, offer h…

• ProEval: Proactive Failure Discovery and Efficient Performance Estimation for Generative AI Evaluation

  Yizheng Huang, Wenjun Zeng, Aditi Kumaresan, Zi Wang · 28 de abril de 2026

  Evaluating generative AI models is increasingly resource-intensive due to slow inference, expensive raters, and a rapidly growing landscape of models and benchmarks. We propose ProEval, a proactive evaluation framework that leverages transfer learning to efficiently estimate performance and identify…

• FastAT Benchmark: A Comprehensive Framework for Fair Evaluation of Fast Adversarial Training Methods

  Chao Pan, Xin Yao · 28 de abril de 2026

  Fast Adversarial Training (FastAT) seeks to achieve adversarial robustness at a fraction of the computational cost incurred by standard multi-step methods such as PGD-AT. Although numerous FastAT techniques have been proposed in recent years, fair comparison among them remains elusive. Existing benc…

• Adversary-Free Counterfactual Prediction via Information-Regularized Representations

  Shiqin Tang, Rong Feng, Shuxin Zhuang, Youzhi Zhang, Hongzong Li · 28 de abril de 2026

  We study counterfactual prediction under assignment bias and propose a mathematically grounded, information-theoretic approach that removes treatment-covariate dependence without adversarial training. Starting from a bound that links the counterfactual-factual risk gap to mutual information, we lear…

• UGAF-ITS: A Standards Harmonization Framework and Validation Tool for Multi-Framework AI Governance in Distributed Intelligent Transportation Systems

  Talal Ashraf Butt, Muhammad Iqbal, Razi Iqbal · 28 de abril de 2026

  Organizations deploying AI-enabled Intelligent Transportation Systems face fragmented governance: ISO/IEC 42001 demands a certifiable management system, the EU AI Act imposes binding high-risk obligations from August 2026, and the NIST AI Risk Management Framework structures voluntary practice. Each…

• Verifying Quantized GNNs With Readout Is Decidable But Highly Intractable

  Artem Chernobrovkin, Marco S\"alzer, Fran\c{c}ois Schwarzentruber, Nicolas Troquard · 28 de abril de 2026

  We introduce a logical language for reasoning about quantized aggregate-combine graph neural networks with global readout (ACR-GNNs). We provide a logical characterization and use it to prove that verification tasks for quantized GNNs with readout are (co)NEXPTIME-complete. This result implies that …

• UniAda: Universal Adaptive Multi-objective Adversarial Attack for End-to-End Autonomous Driving Systems

  Jingyu Zhang, Jacky Wai Keung, Yan Xiao, Yihan Liao, Yishu Li, Xiaoxue Ma · 28 de abril de 2026

  Adversarial attacks play a pivotal role in testing and improving the reliability of deep learning (DL) systems. Existing literature has demonstrated that subtle perturbations to the input can elicit erroneous outcomes, thereby substantially compromising the security of DL systems. This has emerged a…

• Certified geometric robustness -- Super-DeepG

  No\'emie Cohen (Airbus CR\&T), M\'elanie Ducoffe (Airbus CR\&T), Christophe Gabreau, Claire Pagetti, Xavier Pucel · 28 de abril de 2026

  Safety-critical applications are required to perform as expected in normal operations. Image processing functions are often required to be insensitive to small geometric perturbations such as rotation, scaling, shearing or translation. This paper addresses the formal verification of neural networks …

• Unstable Rankings in Bayesian Deep Learning Evaluation

  Qishi Zhan, Minxuan Hu, Guansu Wang, Jiaxin Liu, Liang He · 28 de abril de 2026

  Standard evaluations of Bayesian deep learning methods assume that metric estimates are reliable, but we show this assumption fails under data scarcity. Method rankings are not only unreliable at small $n$, but also dataset-dependent in ways that point estimates cannot reveal: the same method compar…

• Mitigating Error Amplification in Fast Adversarial Training

  Mengnan Zhao, Lihe Zhang, Bo Wang, Tianhang Zheng, Hong Zhong, Geyong Min · 28 de abril de 2026

  Fast Adversarial Training (FAT) has proven effective in enhancing model robustness by encouraging networks to learn perturbation-invariant representations. However, FAT often suffers from catastrophic overfitting (CO), where the model overfits to the training attack and fails to generalize to unseen…

• When Does LLM Self-Correction Help? A Control-Theoretic Markov Diagnostic and Verify-First Intervention

  Aofan Liu, Jingxiang Meng · 27 de abril de 2026

  Iterative self-correction is widely used in agentic LLM systems, but when repeated refinement helps versus hurts remains unclear. We frame self-correction as a cybernetic feedback loop in which the same language model serves as both controller and plant, and use a two-state Markov model over {Correc…

• On the Properties of Feature Attribution for Supervised Contrastive Learning

  Leonardo Arrighi, Julia Eva Belloni, Aur\'elie Gallet, Ivan Gentile, Matteo Lippi, Marco Zullich · 27 de abril de 2026

  Most Neural Networks (NNs) for classification are trained using Cross-Entropy as a loss function. This approach requires the model to have an explicit classification layer. However, there exist alternative approaches, such as Contrastive Learning (CL). Instead of explicitly operating a classificatio…

• Sovereign Agentic Loops: Decoupling AI Reasoning from Execution in Real-World Systems

  Jun He, Deying Yu · 27 de abril de 2026

  Large language model (LLM) agents increasingly issue API calls that mutate real systems, yet many current architectures pass stochastic model outputs directly to execution layers. We argue that this coupling creates a safety risk because model correctness, context awareness, and alignment cannot be …

• PermaFrost-Attack: Stealth Pretraining Seeding(SPS) for planting Logic Landmines During LLM Training

  Harsh Kumar, Rahul Maity, Tanmay Joshi, Aman Chadha, Vinija Jain, Suranjana Trivedy, Amitava Das · 27 de abril de 2026

  Aligned large language models(LLMs) remain vulnerable to adversarial manipulation, and their dependence on web-scale pretraining creates a subtle but serious attack surface. We study Stealth Pretraining Seeding (SPS), a new attack family in which adversaries distribute small amounts of poisoned cont…

• Adversarial Co-Evolution of Malware and Detection Models: A Bilevel Optimization Perspective

  Olha Jure\v{c}kov\'a, Martin Jure\v{c}ek, Matou\v{s} Koz\'ak, R\'obert L\'orencz · 27 de abril de 2026

  Machine learning-based malware detectors are increasingly vulnerable to adversarial examples. Traditional defenses, such as one-shot adversarial training, often fail against adaptive attackers who use reinforcement learning to bypass detection. This paper proposes a robust defense framework based on…

• PoLO: Proof-of-Learning and Proof-of-Ownership at Once with Chained Watermarking

  Haiyu Deng, Yanna Jiang, Guangsheng Yu, Qin Wang, Xu Wang, Baihe Ma, Wei Ni, Ren Ping Liu · 27 de abril de 2026

  Our evaluation shows that PoLO achieves \textbf{99\%} watermark detection accuracy for ownership verification, while preserving data privacy and cutting verification costs to just \textbf{1.5--10\%} of traditional methods. Forging PoLO demands \textbf{1.1--4$\times$} more resources than honest proof…

• ArmSSL: Adversarial Robust Black-Box Watermarking for Self-Supervised Learning Pre-trained Encoders

  Yongqi Jiang, Yansong Gao, Boyu Kuang, Chunyi Zhou, Anmin Fu, Liquan Chen · 27 de abril de 2026

  Self-supervised learning (SSL) encoders are invaluable intellectual property (IP). However, no existing SSL watermarking for IP protection can concurrently satisfy the following two practical requirements: (1) provide ownership verification capability under black-box suspect model access once the st…

• Do Not Imitate, Reinforce: Iterative Classification via Belief Refinement

  Mahdi Kallel, Johannes T\"olle, Ahmed Hendawy, Carlo D'Eramo · 27 de abril de 2026

  Standard supervised classification trains models to imitate the exact labels provided by a perfect oracle. This imitation happens in a single pass, restricting the model to a fixed compute budget even when inputs vary in complexity. Moreover, the rigid training objective forces the model to express …

• Useful nonrobust features are ubiquitous in biomedical images

  Coenraad Mouton, Randle Rabe, Niklas C. Koser, Nicolai Krekiehn, Christopher Hansen, Jan-Bernd H\"ovener, Claus-C. Gl\"uer · 27 de abril de 2026

  We study whether deep networks for medical imaging learn useful nonrobust features - predictive input patterns that are not human interpretable and highly susceptible to small adversarial perturbations - and how these features impact test performance. We show that models trained only on nonrobust fe…

• Estimating Tail Risks in Language Model Output Distributions

  Rico Angell, Raghav Singhal, Zachary Horvitz, Zhou Yu, Rajesh Ranganath, Kathleen McKeown, He He · 27 de abril de 2026

  Language models are increasingly capable and are being rapidly deployed on a population-level scale. As a result, the safety of these models is increasingly high-stakes. Fortunately, advances in alignment have significantly reduced the likelihood of harmful model outputs. However, when models are qu…

• Ethics Testing: Proactive Identification of Generative AI System Harms

  Shin Hwei Tan, Haibo Wang, Heng Li · 27 de abril de 2026

  Generative Artificial Intelligence (GAI) systems that can automatically generate content in the form of source code or other contents (e.g., images) has seen increasing popularity due to the emergence of tools such as ChatGPT which rely on Large Language Models (LLMs). Misuse of the automatically ge…

• Preserve Support, Not Correspondence: Dynamic Routing for Offline Reinforcement Learning

  Zhancun Mu, Guangyu Zhao, Yiwu Zhong, Chi Zhang · 27 de abril de 2026

  One-step offline RL actors are attractive because they avoid backpropagating through long iterative samplers and keep inference cheap, but they still have to improve under a critic without drifting away from actions that the dataset can support. In recent one-step extraction pipelines, a strong iter…