Wireless communication is widely adopted and application oriented technology There are a huge literature about Mobile Ad-hoc NETwork (MANET) is available. In these studies, the ad-hoc network has two major issues security and performance. In this paper, a feasible and adoptable solution is introduced for enhancing security in MANET. The presented paper utilizes the network characteristics and their behavioral difference during attack. Using the attack and normal network behavior a machine learning algorithm is trained and the malicious patterns are distinguished according to the new network samples.