An unstructured version of a traditional wireless network provides a mobile ad hoc network (MANET) that is well suited for emergencies. However, due to lack of infrastructure and resource limitations, many performance problems occur at the same time. High-security vulnerabilities are likely to arise, largely because of dynamic behavior, and the complete communication cycle dependent on unidentified nodes which join and leaves as they want. Several past security schemes studies suggested that reliable and trusted nodes can minimize communication overhead and achieve higher throughput. This kind of scheme affects the route chosen for communication because it cannot distinguish between intentional and unintentional errors at runtime. It is, therefore, important to categorize node activities and also provide a node recovery scheme to regain their original reliability for the genuine nodes. In this paper, we propose a Reliable Trustworthy Approach (RTA) based on node behavior prediction for secure routing. It presents a behavior prediction algorithm for recognizing node behavior by differentiating among the unintentional transient errors and intentional malicious behavior, and the RTA mechanism which provides a method for node trust computation and a method for trust recovery. Experiment results show throughput improvisation against the malicious node intrusion. With increasing, malicious nodes in network RTA mechanism effectively predicted with an optimal packet loss and delay it attained prove the effectiveness of the mechanism.