Timezone: »

 
Poster
Certified Defenses for Data Poisoning Attacks
Jacob Steinhardt · Pang Wei Koh · Percy Liang

Tue Dec 05 06:30 PM -- 10:30 PM (PST) @ Pacific Ballroom #68
in Posters Tue »

Machine learning systems trained on user-provided data are susceptible to data poisoning attacks, whereby malicious users inject false training data with the aim of corrupting the learned model. While recent work has proposed a number of attacks and defenses, little is understood about the worst-case loss of a defense in the face of a determined attacker. We address this by constructing approximate upper bounds on the loss across a broad family of attacks, for defenders that first perform outlier removal followed by empirical risk minimization. Our approximation relies on two assumptions: (1) that the dataset is large enough for statistical concentration between train and test error to hold, and (2) that outliers within the clean (non-poisoned) data do not have a strong effect on the model. Our bound comes paired with a candidate attack that often nearly matches the upper bound, giving us a powerful tool for quickly assessing defenses on a given dataset. Empirically, we find that even under a simple defense, the MNIST-1-7 and Dogfish datasets are resilient to attack, while in contrast the IMDB sentiment dataset can be driven from 12% to 23% test error by adding only 3% poisoned data.

Keywords: Online Learning · Privacy, Anonymity, and Security
[ Code]  [ Paper] 

Author Information

Jacob Steinhardt (Stanford University)
Pang Wei Koh (Stanford University)
Percy Liang (Stanford University)

More from the Same Authors