AI Insight
IDSISA is a new machine learning method that efficiently removes specific training data from AI models to comply with privacy regulations. The approach uses influence functions to group data likely to be deleted into fewer partitions and employs knowledge distillation to maintain model performance, achieving 14-91% better selective forgetting and up to 50% better utility retention compared to the existing SISA method while maintaining formal deletion guarantees. The method was tested across multiple neural network architectures including ResNet-18, BiLSTM, and temporal convolutional networks on standard image and text datasets.
Why it matters
This work addresses a critical challenge in AI privacy compliance, particularly for GDPR's "right to be forgotten," by making it more practical to remove individual user data from trained models without completely retraining them. The approach could enable organizations to handle data deletion requests more efficiently while maintaining model quality and providing formal guarantees about data removal.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Machine unlearning removes the influence of specific training examples to comply with privacy regulations. Full retraining gives strong guarantees but is computationally prohibitive at scale; approximate methods are efficient but lack formal guarantees. SISA (Sharded, Isolated, Sliced, and Aggregated training) achieves certified deletion through data sharding and independent retraining, but random partitioning scatters forget requests across most shards, and isolated retraining degrades utility. We introduce IDSISA (Influence-Distilled SISA), which uses influence functions to concentrate forget-prone samples into few shards and distillation from unaffected shards to recover utility lost to isolation. The deletion guarantee inherited from SISA’s retraining step remains exact and requires no probabilistic qualification; the distillation-based utility recovery is a separately quantified, empirically bounded approximation ($epsilon_{text{cert}} leq 0.015$ in expectation). We further formalize certification under sequences of repeated deletion requests, where cross-shard distillation can create dependencies between shards; we prove that a closure-based cascading retraining procedure restores the deletion guarantee to the full aggregated model under any deletion sequence. Experiments on ResNet-18/CIFAR-10, ResNet-18/CIFAR-100, BiLSTM/IMDB, TCN/IMDB, and a fully connected network on Fashion-MNIST show $14-91%$ better selective forgetting, up to $50%$ better utility retention in relative terms (with consistent percentage-point gains on the other architectures, where a relative baseline is undefined), and unlearning-time changes ranging from $-6.7%$ to $+6.4%$ depending on architecture, relative to SISA.
Source: IDSISA: Influence-Distilled Sharded Training for Efficient Machine Unlearning