Generally an agent-based security system is constructed hierarchical, with a central manager acting as head of the whole system. However, the manager becomes a bottleneck for being connected by each client. It can even overload when too many clients request service simultaneously. The whole system may collapse when the central manager is attacked. And these systems are passive to detect and deal with the secure problem. Hereby we present a mobile agent-based P2P Model for Autonomous Security Hole Discovery (MASHD). It can detect infection and network intrusion based on knowledge of the local host. Viruses will be removed and connection will be refused after identification. In case of a suspicious activity, MASHD initiates a voting approach to make a collective decision and take further action. The integration of peer-to-peer behavior with mobile agents reduces latency and load; however, flexibility, effectivity, security and cooperation of the system are enhanced.