Improving host-based IDS with argument abstraction to prevent mimicry attacks

  • Authors:
  • Sufatrio;Roland H. C. Yap

  • Affiliations:
  • Temasek Laboratories, National University of Singapore, Singapore, Singapore;School of Computing, National University of Singapore, Singapore, Singapore

  • Venue:
  • RAID'05 Proceedings of the 8th international conference on Recent Advances in Intrusion Detection
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

A popular class of host-based Intrusion Detection Systems (IDS) are those based on comparing the system call trace of a process against a set of k-grams. However, the detection mechanism in such IDS can be evaded by cloaking an attack as a mimicry attack. In this paper, we give an algorithm that transforms a detectable attack into a mimicry attack. We demonstrate on a number of examples that using this algorithm, mimicry attacks can be easily constructed on self-based IDS with a set of k-grams and also a more precise graph profile representation. We enhance the IDS by making use of the system call arguments and process credentials. To avoid increasing the false positives, a supplied specification is used to abstract the system call arguments and process credentials. The specification takes into account what objects in the system that can be sensitive to potential attacks, and highlights the occurrence of “dangerous” operations. With this simple extension, we show that the robustness of the IDS is increased. Our preliminary experiments show that on our example programs and attacks, it was no longer possible to construct mimicry attacks. We also demonstrate that the enhanced IDS provides resistance to a variety of common attack strategies.