MANETs are in-secure and vulnerable to attacks, as it lacks a central trusted authority. Providing security to such a network becomes an essential task and formulate the origin of the proposed work. The proposed system attempts to secure MANETs using route validation and cryptographic techniques. One of the most crucial, and primary concerns in today’s times is to be able to detect an attacker at its initial stages. With this focus point in mind, the approach used is to prevent such an attacks by detecting in initial stage and preventing from network degradation. The novelty of the proposal is the use of cryptographic techniques for improving the security, along a reverse-AODV for reducing path fail correction, and machine learning concepts for validation of results. Many of the existing malware detection techniques proposed by the researchers were executed either on machine independent platform, or on an available dataset with machine dependent approaches. This drawback has been addressed in the existing proposal, where in machine learning is used with self-generated data-set, to eliminate contingent problems. The proposed system includes a network that is free from malware. Justification of the results were generated using classifier tools that were trained with the obtained dataset. For secured communication, an Elliptical curve cryptographic algorithm is applied to reverse ad-hoc on-demand distance vector with reverse multiple route replies that have been generated from the destination to the source node. An investigation to ensure the correct delivery of data can be done by diverting the traffic through the shortest alternative secured path. The metrics used for statistical analysis include average transmission-delay, overhead, packet forwarding rate and packet- delivery-rate, based on which the conclusion is theorized for the detection. The major finding includes the process of selecting an appropriate condition for detecting a malicious node, observing the network behavior with varying number of suspicious nodes and then validating the correctness. The implementation gives us varied results, both when the suspicious node is deferred for some time, and then on its complete elimination. The limitation of the proposal is that the suspicious nodes are uncoordinated.