This article focused on the detection techniques of malware in network level. The utilization of proposed technique try to detect most of the metamorphic malware packets that transmitted in the network. Besides, this study has focused on analyzing the malware features to predict their pattern in the network using different types of the classifiers such as Naïve bias, neural network, and Support vector machine (SVM) classifier. Various software and tools have been employed in this study such as Weka, Wireshark, Ida-pro, Matlab…etc. The proposed technique that used in this study is concentrated on accelerating the normal packets which transferred in network without needing to send them to slow path by using proposed middle path and also to detect many malware packets.