We review an architecture for intrusion detection based on autonomous agents, which can avoid the single point failure of the distributed systems, balance the system load , and improve the efficiency to meet the requirements of modern computer and network technologies.