{"ID":2854420,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.14299","arxiv_id":"2510.14299","title":"TED++: Submanifold-Aware Backdoor Detection via Layerwise Tubular-Neighbourhood Screening","abstract":"As deep neural networks power increasingly critical applications, stealthy backdoor attacks, where poisoned training inputs trigger malicious model behaviour while appearing benign, pose a severe security risk. Many existing defences are vulnerable when attackers exploit subtle distance-based anomalies or when clean examples are scarce. To meet this challenge, we introduce TED++, a submanifold-aware framework that effectively detects subtle backdoors that evade existing defences. TED++ begins by constructing a tubular neighbourhood around each class's hidden-feature manifold, estimating its local ``thickness'' from a handful of clean activations. It then applies Locally Adaptive Ranking (LAR) to detect any activation that drifts outside the admissible tube. By aggregating these LAR-adjusted ranks across all layers, TED++ captures how faithfully an input remains on the evolving class submanifolds. Based on such characteristic ``tube-constrained'' behaviour, TED++ flags inputs whose LAR-based ranking sequences deviate significantly. Extensive experiments are conducted on benchmark datasets and tasks, demonstrating that TED++ achieves state-of-the-art detection performance under both adaptive-attack and limited-data scenarios. Remarkably, even with only five held-out examples per class, TED++ still delivers near-perfect detection, achieving gains of up to 14\\% in AUROC over the next-best method. The code is publicly available at https://github.com/namle-w/TEDpp.","short_abstract":"As deep neural networks power increasingly critical applications, stealthy backdoor attacks, where poisoned training inputs trigger malicious model behaviour while appearing benign, pose a severe security risk. Many existing defences are vulnerable when attackers exploit subtle distance-based anomalies or when clean ex...","url_abs":"https://arxiv.org/abs/2510.14299","url_pdf":"https://arxiv.org/pdf/2510.14299v1","authors":"[\"Nam Le\",\"Leo Yu Zhang\",\"Kewen Liao\",\"Shirui Pan\",\"Wei Luo\"]","published":"2025-10-16T04:51:25Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.AI\"]","methods":"[]","has_code":false,"code_links":[{"ID":608152,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2854420,"paper_url":"https://arxiv.org/abs/2510.14299","paper_title":"TED++: Submanifold-Aware Backdoor Detection via Layerwise Tubular-Neighbourhood Screening","repo_url":"https://github.com/namle-w/TEDpp","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}