Software integrity protection using timed executable agents Conference Paper uri icon


  • We present a software scheme for protecting the integrity of computing platforms using Timed Executable Agent Systems (TEAS). A trusted challenger issues an authenticated challenge to a perhaps corrupt responder. New is that the issued challenge is an executable program that can potentially compute any function on the responder. The responder must compute not only the correct value implied by the agent, but also must complete this computation within time bounds prescribed by the challenger. Software-based attestation schemes have been proposed before - now capabilities introduced in TEAS provide means to mitigate the existing shortcomings of such proposed techniques. TEAS are general and can be adapted to many applications for which system integrity is to be tested. Two types of adversaries to TEAS are considered. First, we address attacks by "off-line" adversaries that attempt to discern agents' functions statically by analyzing their program texts. We then consider "on-line" adversaries, which operate while the agent runs. For off-line adversaries, we show how complexity results from programming language analysis, as well as undecidability considerations, can be used to thwart such analysis by making it impossible for the adversary to correctly decipher all potential agents and reply in a timely fashion. In the on-line scenario, adversaries are difficult to stop in general. We do however present strategics that make it difficult for these adversaries to interpret an agent in a virtual machine and to thereby redirect its actions, for example. We address the problem of creating large libraries of useful and complicated (and hence difficult to analyze) agents through a new technique of program blinding - we hide critical functionality inside randomly generated machine-language programs. We implemented a virtual machine that allows experimentation with this approach. Experiments reveal that blinded agents whose execution conveys important integrity information can be efficiently generated in abundance. Copyright 2006 ACM.

name of conference

  • the 2006 ACM Symposium

published proceedings

  • Proceedings of the 2006 ACM Symposium on Information, computer and communications security - ASIACCS '06

author list (cited authors)

  • Garay, J. A., & Huelsbergen, L.

citation count

  • 20

complete list of authors

  • Garay, Juan A||Huelsbergen, Lorenz

publication date

  • January 2006