Data space randomization for securing cyber-physical systems

Bradley Potteiger, Feiyang Cai, Zhenkai Zhang, Xenofon Koutsoukos

Research output: Contribution to journalArticlepeer-review

Abstract

Non-control data attacks have become widely popular for circumventing authentication mechanisms in websites, servers, and personal computers. These attacks can be executed against cyber-physical systems (CPSs) in which not only authentication is an issue, but safety is at risk. Furthermore, any unauthorized change to safety-critical variables within the software may cause damage or even catastrophic consequences. Moving target defense techniques such as data space randomization (DSR) have become popular for protecting against memory corruption attacks such as non-control data attacks. However, current DSR implementations rely on source code transformations and do not stop critical variables from being overwritten, only that the new overwritten value will be vastly different than expected by the attacker. As such, these implementations are often ineffective for legacy CPS software in which only a binary is available. The problem addressed in this paper is how do we protect against non-control data attacks in legacy CPS software while ensuring that we can detect instances of variable integrity violations. We solve this problem by combining DSR at the binary level with variable comparison checks to ensure that we can detect and mitigate any attacker attempt to overwrite safety-critical variables. Our security approach is demonstrated utilizing an autonomous emergency braking system case study.

Original languageEnglish
JournalInternational Journal of Information Security
DOIs
StateAccepted/In press - 2021

Keywords

  • Autonomous vehicles
  • Cyber-physical systems
  • Data space randomization
  • Moving target defenses
  • Resiliency

Fingerprint

Dive into the research topics of 'Data space randomization for securing cyber-physical systems'. Together they form a unique fingerprint.

Cite this