Runtime Integrity Checking for Exploit Mitigation on Lightweight Embedded Devices

Authors

Matthias Neugschwandtner, Collin Mulliner, William Robertson, Engin Kirda

Venue

Proceedings of the 9th International Conference on Trust & Trustworthy Computing (TRUST), August 2016

Abstract

Entering the age of the Internet of things, embedded devices are everywhere. They are built using common hardware such as RISC-based ARM and MIPS platforms, and lightweight open software components. Because of their limited resources, such systems often lack the protection mechanisms that have been introduced to the desktop and server world. In this paper, we present BINtegrity, a novel approach for exploit mitigation that is specifically tailored towards embedded systems that are based on the common RISC architecture. BINtegrity leverages architectural features of RISC CPUs to extract a combination of static and dynamic properties relevant to OS service requests from executables, and enforces them during runtime. Our technique borrows ideas from several areas including system call monitoring, static analysis, and code emulation, and combines them in a low-overhead fashion directly in the operating system kernel. We implemented BINtegrity for the Linux operating system. BINtegrity is practical, and restricts the ability of attackers to exploit generic memory corruption vulner- abilities in COTS binaries. In contrast to other approaches, BINtegrity does not require access to source code, binary modification, or application specific configuration such as policies. Our evaluation demonstrates that BINtegrity incurs a very low overhead – only 2%, – and shows that our approach mitigates both code injection and code reuse attacks.

BibTeX

@inproceedings{Neugschwandtner2016Runtime_Integrity,
  title     = {{Runtime Integrity Checking for Exploit Mitigation on Lightweight Embedded Devices}},
  author    = {Neugschwandtner, Matthias and Mulliner, Collin and Robertson, William and Kirda, Engin},
  booktitle = {Proceedings of the 9th International Conference on Trust \& Trustworthy Computing},
  series    = {TRUST},
  month     = {August},
  year      = {2016}
}