Selective Amnesia: On Efficient, High-Fidelity and Blind Suppression of Backdoor Effects in Trojaned Machine Learning Models
This addresses the security issue of backdoor attacks in ML models for practitioners, offering a fast and data-efficient solution, though it is incremental as it builds on continual learning concepts.
The paper tackles the problem of removing backdoors from Trojaned machine learning models by introducing SEAM, a technique that induces selective amnesia through random-label retraining and recovery, achieving high fidelity and efficiency with a 30x speedup over retraining from scratch on MNIST and using only 0.1% of clean data for TrojAI models.
In this paper, we present a simple yet surprisingly effective technique to induce "selective amnesia" on a backdoored model. Our approach, called SEAM, has been inspired by the problem of catastrophic forgetting (CF), a long standing issue in continual learning. Our idea is to retrain a given DNN model on randomly labeled clean data, to induce a CF on the model, leading to a sudden forget on both primary and backdoor tasks; then we recover the primary task by retraining the randomized model on correctly labeled clean data. We analyzed SEAM by modeling the unlearning process as continual learning and further approximating a DNN using Neural Tangent Kernel for measuring CF. Our analysis shows that our random-labeling approach actually maximizes the CF on an unknown backdoor in the absence of triggered inputs, and also preserves some feature extraction in the network to enable a fast revival of the primary task. We further evaluated SEAM on both image processing and Natural Language Processing tasks, under both data contamination and training manipulation attacks, over thousands of models either trained on popular image datasets or provided by the TrojAI competition. Our experiments show that SEAM vastly outperforms the state-of-the-art unlearning techniques, achieving a high Fidelity (measuring the gap between the accuracy of the primary task and that of the backdoor) within a few minutes (about 30 times faster than training a model from scratch using the MNIST dataset), with only a small amount of clean data (0.1% of training data for TrojAI models).