AI & ML
SCR: Detecting Backdoor-Poisoned Samples through Clean-Reference Residuals
arXiv cs.CR0 views
arXiv:2611.24558v1 Announce Type: cross Abstract: Deep neural networks are vulnerable to backdoor attacks, in which adversaries manipulate model predictions by injecting poisoned samples into training data. Detecting these samples remains challenging because legitimate semantic variation can obscure poisoning-related signals in learned representations. Consequently, class-level statistics used by existing detectors may flag benign feature deviations while overlooking poisoned samples that blend into the class distribution. In this paper, we propose Semantic Counterfactual Residualization (SCR), which uses a small set of clean reference samples to account for benign variation through sample-specific feature reconstruction. SCR identifies poisoned samples by detecting coherent deviations from these clean baselines, without requiring prior knowledge of the attacks. We evaluate SCR through experiments covering twelve dirty-label and clean-label attacks across CIFAR-10, CIFAR-100, and Tiny ImageNet. With a small clean-reference set, SCR achieves attack-averaged true positive rates above 99.5\% and false positive rates below 2\% on each dataset, outperforming eight representative detection baselines in macro-TPR.
Source:https://arxiv.org/abs/2611.24558PreviousRECLAIM: Robot-Vision-Informed Adversarial Patch Defense via Embodied Attention FeedbackNext"I Don't Think It's a Scam": A First Look at Family Intervention Barriers in Elder Abuse Discussions on Reddit and Baidu Tieba
Related posts
Gig Work Is Invading Nursing Homes. The Risks Can Be Deadly
0Wired
4 Request-Validation Gotchas When Migrating OpenAI Clients to GPT-6.2 Reasoning Models
0dev.to
Razorpay webhook not verified: anyone can fake a payment
0dev.to
From FAQPage Schema to Answerable AI Documentation
0dev.to
Wiring Android's MediaPipe LLM Inference API to a Multi-Turn Conversation Engine
0dev.to
What Is LLM Throughput? Tokens per Second, Concurrency, and Batching
0dev.to