Detecting and Proving Manipulation Attacks in Mobile Agent Systems

Mobile agents are software entities consisting of code, data and state that can migrate autonomously from host to host executing their code. Unfortunately, security issues restrict the use of mobile agents despite its benefits. The protection of mobile agents against the attacks of malicious hosts is considered the most difficult security problem to solve in mobile agent systems. In a previous work, the Mobile Agent Watermarking approach (MAW) was presented as a new attack detection technique based on embedding a fixed watermark into the agent’s code. In this paper, some improvements are introduced in MAW. Instead of a fixed watermark, the origin host embeds a watermark that can change dynamically during execution. In each host, the marked code creates a data container where the watermark will be transferred and the results will be hidden. When the agent returns home, the origin host applies a set of integrity rules that the containers must fulfill. These rules can be inferred from the modifications performed in the agent’s code during the watermark embedding. If a container does not fulfill the rules, this means that the corresponding host is malicious. This paper also presents how these containers can be used as a proof to demonstrate that a manipulation attack has been performed.