Optimal Adversarial Attack on Autoregressive Models

02/01/2019
by   Yiding Chen, et al.
0

We investigate optimal adversarial attacks against time series forecast made by autoregressive models. In our setting, the environment evolves according to a potentially nonlinear dynamical system. A linear autoregressive model observes the current environment state and predicts its future values. But an adversary can modify the environment state and hence indirectly manipulate the autoregressive model forecasts. The adversary wants to drive the forecasts towards some adversarial targets while minimizing environment modification. We pose this attack problem as optimal control. When the environment dynamics is linear, we provide a closed-form solution to the optimal attack using Linear Quadratic Regulator (LQR). Otherwise, we propose an approximate attack based on Model Predictive Control (MPC) and iterative LQR (iLQR). Our paper thus connects adversarial learning with control theory. We demonstrate the advantage of our methods empirically.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
11/11/2018

An Optimal Control View of Adversarial Machine Learning

I describe an optimal control view of adversarial machine learning, wher...
research
03/05/2019

Online Data Poisoning Attack

We study data poisoning attacks in the online learning setting where the...
research
05/23/2019

Robust guarantees for learning an autoregressive filter

The optimal predictor for a linear dynamical system (with hidden state a...
research
10/23/2022

LQGNet: Hybrid Model-Based and Data-Driven Linear Quadratic Stochastic Control

Stochastic control deals with finding an optimal control signal for a dy...
research
11/20/2022

Adversarial Cheap Talk

Adversarial attacks in reinforcement learning (RL) often assume highly-p...
research
02/29/2020

Optimal Feature Manipulation Attacks Against Linear Regression

In this paper, we investigate how to manipulate the coefficients obtaine...
research
10/30/2020

Adversarial Attacks on Optimization based Planners

Trajectory planning is a key piece in the algorithmic architecture of a ...

Please sign up or login with your details

Forgot password? Click here to reset