Indiscriminate Poisoning Attacks on Unsupervised Contrastive Learning
This addresses security vulnerabilities in contrastive learning for machine learning practitioners, though it is incremental as it extends known poisoning attacks to a new context.
This paper tackles the problem of indiscriminate data poisoning attacks on unsupervised contrastive learning, showing that their proposed Contrastive Poisoning attack drastically reduces performance and is highly generalizable to supervised models.
Indiscriminate data poisoning attacks are quite effective against supervised learning. However, not much is known about their impact on unsupervised contrastive learning (CL). This paper is the first to consider indiscriminate poisoning attacks of contrastive learning. We propose Contrastive Poisoning (CP), the first effective such attack on CL. We empirically show that Contrastive Poisoning, not only drastically reduces the performance of CL algorithms, but also attacks supervised learning models, making it the most generalizable indiscriminate poisoning attack. We also show that CL algorithms with a momentum encoder are more robust to indiscriminate poisoning, and propose a new countermeasure based on matrix completion. Code is available at: https://github.com/kaiwenzha/contrastive-poisoning.