SAGE: Similarity-Based Cleaning of Poisoned Training Data from Verified Examples

Excerpt

arXiv:2610.01788v1 Announce Type: new Abstract: As machine learning increasingly relies on public, untrusted data sources, data poisoning attacks, which inject malicious examples into training data to induce misclassification of a chosen target, pose a growing threat. Existing defenses either assume zero ground-truth information about which examples are poisoned, or they assume access to a large set of examples verified to be clean. Satisfying the latter assumption incurs significant cost since