Intrusion Detection Systems (IDSs) are security tools that attempt to detect malicious activities, which are targeted against a network and its resources. As internetworking among computer systems via the internet becomes more widely and keeps rapid increase, widespread attacks involving those networks occurs more frequently which present a new challenge to IDSs. Many approaches have been used in implementing a reliable IDS like neural networks, statistical analysis… etc, but they have some limitations. In this paper we propose a multi-agent based intrusion detection system,
which will satisfy not all but most of the requirements of reliable and secure IDS. The main goals of this system, which distinguish it from other solutions, are its distributed architecture, scalability, efficiency and the use of Agent concept.