NetDetector: an Anomaly Detection Platform for Networked Systems

Network is an essential infrastructure for mobile edge clouds. The communication between mobile devices and the communication between mobile device and the cloud both rely on the network infrastructure. The reliability of network infrastructure is important to ensure the QoS (Quality of Services) of mobile edge clouds. However, the current network suffers from many attacks. Early detect network attacks is very important. Traditional anomaly network detection methods have some weaknesses, such as slow detection speed, fail to recognize new anomalies, and insufficient accuracy. In order to solve these problems, in this paper we design and implement a new anomaly detection platform - NetDetector for networked systems. We first present the architecture design and then show the implementation details. NetDetector consists of five important modules, and its detection algorithm is based on LSTM (Long Short Term Memory), a special kind of recurrent neural network. A case study is proposed to show the detection performance of NetDetector. Experimental results show NetDetector achieves about 8% performance improvement as compared with the systems that use traditional machine learning methods.