
TL;DR
This paper demonstrates that training-data poisoning can subtly and effectively cause targeted misclassification in acoustic neural networks without affecting overall accuracy, and proposes cryptographic defenses to verify data integrity.
Contribution
It introduces a novel poisoning attack on acoustic classifiers, proves its stealth based on class imbalance, and proposes a cryptographic defense framework for data provenance.
Findings
High attack success rate with minimal data corruption
Aggregate accuracy remains stable despite targeted attacks
Cryptographic methods can verify data integrity
Abstract
Training-data poisoning attacks can induce targeted, undetectable failure in deep neural networks by corrupting a vanishingly small fraction of training labels. We demonstrate this on acoustic vehicle classification using the MELAUDIS urban intersection dataset (approx. 9,600 audio clips, 6 classes): a compact 2-D convolutional neural network (CNN) trained on log-mel spectrograms achieves 95.7% Attack Success Rate (ASR) -- the fraction of target-class test samples misclassified under the attack -- on a Truck-to-Car label-flipping attack at just p=0.5% corruption (48 records), with zero detectable change in aggregate accuracy (87.6% baseline; 95% CI: 88-100%, n=3 seeds). We prove this stealth is structural: the maximum accuracy drop from a complete targeted attack is bounded above by the minority class fraction (beta). For real-world class imbalances (Truck approx. 3%), this bound falls…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdversarial Robustness in Machine Learning · Physical Unclonable Functions (PUFs) and Hardware Security · Advanced Malware Detection Techniques
