Anomaly Detection for Cyber-Security Based on Convolution Neural Network : A survey
Montdher Alabadi, Yüksel Çelik
Abstract
The expanding growth of computer and communication technologies results in a vast amount of security concerns. Various types of cyber-security enabled mechanisms have been developed to limit these concerns. Anomaly detection is among these mechanisms. Anomaly detection means using multiple techniques and methods to detect different patterns that do not conform to defined features of whole data. Recently, deep learning techniques adopted as a satisfactory solution because of its ability to extract data features from data itself. Convolution neural network (CNN) is mainly utilized because of its ability to process input with multi-dimensions. In this paper, a comprehensive survey about using CNN as a key solution for anomaly detection is provided. Most of the existing solutions in the literature have been gathered and classified according to the input data source; furthermore, this paper suggests a unified cross framework that simulates end-to-end anomaly detection mechanisms that exist in the previous studies. A unified cross framework enriches this paper with in-depth analysis to clarify how the solution in the literature uses CNN in anomaly detection. Finally, this paper suggests several future research directions that can support the audience in their future works in this context.