Mobile Agent Route Protection through Hash-Based Mechanisms

Abstract

One approach to secure mobile agent execution is restricting the agent route to trusted environments. A necessary condition for this approach to be practical is that the agent route be protected. Previous proposals for agent route protection either offer low security or suffer from high computational costs due to cryptographic operations. We present two fast, hash-based mechanisms for agent route protection. The first solution relies on hash collisions and focuses on minimizing the computational cost of route verification by hosts along the route; the cost is shifted to the stage of route protection by the agent owner. The second solution uses Merkle trees and minimizes the cost of route protection by the agent owner, so that a single digital signature suffices to protect the whole route; for hosts along the route, the verification cost is similar to the cost of previous schemes in the literature, namely one digital signature verification per route step. The first solution is especially suitable for agent routes which go through heavily loaded hosts (to avoid denial of service or long delay). The second solution is more adapted to mitigating the bottleneck at agent owners who are expected to launch a great deal of agents. Both solutions provide independent protection for each route step and can be extended to handle flexible itineraries.

This work is partly supported by the Spanish CICYT under project no. TEL98- 0699-C02-02.