TruVRF: Towards Triple-Granularity Verification on Machine Unlearning
This addresses the problem of verifying machine unlearning for data contributors and model providers, though it appears incremental as it builds on existing unlearning frameworks like SISA and Amnesiac Unlearning.
The paper tackles the lack of reliable validation methods for machine unlearning, which risks dishonest model providers misleading data contributors, by introducing TruVRF, a non-invasive verification framework that operates at class-, volume-, and sample-level granularities and achieves over 90% accuracy for two metrics and 4.8% to 8.2% inference deviation for another in evaluations on three datasets.
The concept of the right to be forgotten has led to growing interest in machine unlearning, but reliable validation methods are lacking, creating opportunities for dishonest model providers to mislead data contributors. Traditional invasive methods like backdoor injection are not feasible for legacy data. To address this, we introduce TruVRF, a non-invasive unlearning verification framework operating at class-, volume-, and sample-level granularities. TruVRF includes three Unlearning-Metrics designed to detect different types of dishonest servers: Neglecting, Lazy, and Deceiving. Unlearning-Metric-I checks class alignment, Unlearning-Metric-II verifies sample count, and Unlearning-Metric-III confirms specific sample deletion. Evaluations on three datasets show TruVRF's robust performance, with over 90% accuracy for Metrics I and III, and a 4.8% to 8.2% inference deviation for Metric II. TruVRF also demonstrates generalizability and practicality across various conditions and with state-of-the-art unlearning frameworks like SISA and Amnesiac Unlearning.