Online Data Poisoning Attack
arXiv:1903.01666
Abstract
We study data poisoning attacks in the online setting where training items arrive sequentially, and the attacker may perturb the current item to manipulate online learning. Importantly, the attacker has no knowledge of future training items nor the data generating distribution. We formulate online data poisoning attack as a stochastic optimal control problem, and solve it with model predictive control and deep reinforcement learning. We also upper bound the suboptimality suffered by the attacker for not knowing the data generating distribution. Experiments validate our control approach in generating near-optimal attacks on both supervised and unsupervised learning tasks.
References in corpus (11)
- Continuous control with deep reinforcement learning
- Playing Atari with Deep Reinforcement Learning
- Understanding Black-box Predictions via Influence Functions
- Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
- Support Vector Machines under Adversarial Label Contamination
- On the Sample Complexity of the Linear Quadratic Regulator
- Data Poisoning Attacks against Online Learning
- Adversarial Attacks on Stochastic Bandits
- Online Linear Quadratic Control
- An Optimal Control Approach to Sequential Machine Teaching
- An Optimal Control View of Adversarial Machine Learning
Cited by in corpus (5)
- Data Poisoning against Differentially-Private Learners: Attacks and Defenses
- Using Machine Teaching to Investigate Human Assumptions when Teaching Reinforcement Learners
- Gradient-based Data Subversion Attack Against Binary Classifiers
- The Sample Complexity of Teaching-by-Reinforcement on Q-Learning
- Lethean Attack: An Online Data Poisoning Technique