Type: Conference or Workshop Item (Paper)
Endsuleit, Regine and Wagner, Arno
Possible attacks on and countermeasures for secure multi-agent computation.
In: Int. Conference on Security and Management 2004, Las Vegas, Nevada, USA.
Full text not available from this archive.
In this paper we improve the model for secure multi-agent
computation proposed by Endsuleit and Mie. We apply a recent protocol from Hirt and Maurer for secure multi-party computations to build Alliances of $n$ agents that solve a common task. The protocol is very efficient with a communication load of O(n^2* m)(where m is the number of multiplications). All computations within the Alliance are robust as long as not more than t_max:=
n/3-1 agents are corrupted at the same time.
The main contribution of this paper is an analysis of the t_max-limit under realistic network assumptions. We use an attack tree to identify possible attacks and discuss countermeasures. We also analyse concrete examples for framework and Alliance sizes, depending on the number of malicious hosts in the framework. In addition, we discuss different possibilities to improve Alliance security and to mitigate Denial-of-Service (DoS) attacks.
Archive Staff Only: edit this record