Reinforcement learning for adaptive cyber defense against zero-day attacks

Zhisheng Hu, Ping Chen, Minghui Zhu, Peng Liu

Research output: Chapter in Book/Report/Conference proceedingChapter

9 Scopus citations

Abstract

In this chapter, we leverage reinforcement learning as a unified framework to design effective adaptive cyber defenses against zero-day attacks. Reinforcement learning is an integration of control theory and machine learning. A salient feature of reinforcement learning is that it does not require the defender to know critical information of zero-day attacks (e.g., their attack targets, and the locations of the vulnerabilities). This information is difficult, if not impossible, for the defender to gather in advance. The reinforcement learning based schemes are applied to defeat three classes of attacks: strategic attacks where the interactions between an attacker and a defender are modeled as a non-cooperative game; non-strategic random attacks where the attacker chooses its actions by following a predetermined probability distribution; and attacks depicted by Bayesian attack graphs where the attacker exploits combinations of multiple known or zero-day vulnerabilities to compromise machines in a network.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages54-93
Number of pages40
DOIs
StatePublished - 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11830 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Reinforcement learning for adaptive cyber defense against zero-day attacks'. Together they form a unique fingerprint.

Cite this