Reliable Mobile Agent based system for suspicious MANET

Application of Mobile Agents in MANET is growing day by day to solve various network and user related problems in more efficient manner. However, before mobile agent's viable use in MANET, major performance issue such as reliability of Mobile Agent based System needs to pay attention. Reliability of a mobile agent can be defined as its success ratio to fulfill its objective.) MANET may have some malicious hosts which can affect the reliability of Mobile Agent based System (MAS). To overcome this issue, mobile agent should have some mechanism to find out the presence of these malicious hosts in MAS. Owner Host of Mobile Agent prepares the list of hosts the mobile agent required to visit to fulfill its task. This host list needs to be updated to remove the malicious hosts from the host list. Also, selection of suitable migration policy of agents is also necessary to ensure reliable mobile Agent based system. This paper focuses on the enhancement of host list, selection of migration policy and MAS reliability based on Reinforcement Learning. This method helps mobile agent to learn about a preferred policy for selecting a suitable migration pattern and detecting malicious hosts present in the MANET. To measure the reliability of learning Mobile Agents, a Monte Carlo simulation based algorithm is proposed. Obtained results verifies that the learning of mobile agents contributes towards better reliability of MAS in suspicious MANET.