Machine learning attacks target composed cyber physical systems

Data Manipulation Attacks Against Machine Learning Integrated Cyber-Physical Systems

Machine Learning

Summary

Machine learning models are often combined to control real-world devices like robots or smart infrastructure, but it’s unclear how attacks on single models affect the entire system. The authors study how an attacker could manipulate data to trick multiple connected machine learning components working together in these cyber-physical systems. They propose a way to create an attack plan that decides how to tamper with data based on the system’s state. This work aims to start discussions and collaborations about protecting such complex systems from coordinated attacks.

What this means in practice

  • For industrial control teams: Design strategies to defend complex cyber-physical systems from attacks that manipulate multiple interconnected ML components.
  • For autonomous vehicle engineers: Develop testing frameworks that simulate coordinated attacks on layered ML models controlling vehicle behavior to improve resilience.

A position paper. It proposes an approach and reports no results.

Authors

Soyon Choi, Scott Alfeld, Meiyi Ma

Abstract

There is a well-established body of work demonstrating that individual machine learning (ML) algorithms can be manipulated or deceived. However, in many real-world systems, multiple ML algorithms are composed to enable automated decision-making, particularly in Cyber-Physical Systems (CPS). Often such systems use a composition of vision, planning, and other ML models to affect their internal state and ultimate take actions in the physical world. While there is a huge literature on the (in)security of ML algorithms, it is not immediately clear how an attacker might leverage their ability to trick individual learners when attacking the composition as a whole. We develop methods for creating an attack policy (a mapping from state to attack) for a data-manipulating attacker affecting a CPS or other state-machine-based AI compositions. We describe our framework, a basic threat model and corresponding methods for computing an optimal attack policy, as well as planned extensions. The goal of this work-in-progress paper is to stimulate conversation and forge new collaborations regarding the security of compositional learning.