Independent Reading · 2020

Decoding the Imitation Security Game: Handling Attacker Imitative Behavior Deception

Thanh H. Nguyen, Nam Vu, Amulya Yadav, and Uy Nguyen · ECAI

Security GamesDeceptionAI Security

Full text

Read the Paper

Use View PDF or Download PDF to read the full paper on this device.

Paper Information

Authors
Thanh H. Nguyen, Nam Vu, Amulya Yadav, and Uy Nguyen
Venue
ECAI · 2020
Open official paper

Summary

This paper studies attackers that manipulate historical data by imitating a behavioural model, causing a defender to learn an ineffective strategy.

Core Idea

A defender should explicitly account for strategic deception when converting a learned attacker model into a defense policy.

Method

The authors formulate attacker deception as an optimisation problem and propose a game-theoretic counter-deception algorithm for selecting defense strategies.

Key Takeaways

  • Historical attack data may be strategically generated rather than passively observed.
  • Ignoring deception can cause substantial defender losses.
  • Counter-deception policies can reduce the impact of manipulated learning data.

Questions

  • How can counter-deception methods operate when several attackers adapt simultaneously?
  • Which behavioural assumptions are most important for reliable defense planning?