Towards secure mobile agents

The deployment of mobile agents in large scale on the Internet can not succeed without securing agents them selves. In fact, the transfers of agents over the Internet is not secured also hosts visited by mobile agents can be malicious and try to analyse or modify the agent code, data or behaviour. In our researches, we are interesting to secure agents on visited hosts. So, in this paper we will present the state of the art of securing mobile agents against malicious hosts and the approach that we suggest. In such state of the art, we identify strong and high cost approaches based on the use of additional hardware and weak and low cost approaches based on software solutions. In our approach, we replace the additional hardware by a software solution that we call SVM (Secure Virtual Machine). Thus, an agent is ciphered, signed (in part) and time stamped. On visited host, the agent will be executed by the secure virtual machine SVM without decrypting it. The cipher algorithm that we use is based on the segmentation of the agent, obfuscation of the segments, a symmetric encryption of every segment and the mess-up of segment execution sequence. To provide strong security, we will use smart cards to protect SVM itself.