A Novel Deep Learning Framework for Intrusion Detection System
Mahwish Amjad, Hira Zahid, Sania Zafar, Tariq Mahmood · 2020
Rapid increase of network devices have brought several complexities in today's network data. Deep learning algorithms provides better solution for analyzing complex network data. Several deep learning algorithms have been proposed by researchers for identifying either known or unknown intrusions present in network traffic. But, in real time, incoming network traffic might encounter with known or unknown intrusions. Presence of unknown intrusions in network traffic arises a need to bring a framework that can identify both known and unknown network traffic intrusions. This paper is an attempt to bring a novel deep learning framework that can identify both known or unknown attacks with maximum 82% accuracy. Also, the particular category of known attack will be revealed via proposed framework. Proposed framework is a novel integration of two well known deep learning algorithms autoencoder and LSTM that brings an effective intrusion detection system. We believe that deployment of proposed framework in real time network will bring improvement in the security of future internet.