Homa Taherpour Gelsefid - Bachelor Student in Computer Engineering, Faculty of Technology and Engineering, East of Guilan, University ofGuilan, Guilan, Iran
Seyyed Abdorreza Hesam Mohseni - University Lecturer of Computer Engineering, Faculty of Technology and Engineering, East of Guilan, University ofGuilan, Guilan, Iran

This paper delves into examining the utilization of autoencoders in unsupervised deep learning techniques applied to Network-Based Anomaly Intrusion Detection Systems (IDS). Given the inadequacy of anomaly-base traditional IDSs in detecting zero-day attacks, enhancing their performance in that aspect remains an active research pursuit. This study conducts a comprehensive review of two Denoising Autoencoder (DAE) and sparse autoencoder approaches for identifying novel attacks. The models utilizing AE aim to generate distinctive features conducive to detecting network intrusions. By considering either the number of citations or the significance of emerging methods, relevant works were identified, thoroughly examined, and summarized. The cybersecurity datasets employed in this investigation are publicly accessible and widely recognized. Furthermore, the primary focus of this study is on various autoencoder methodologies within self-taught learning, serving as an automated means for feature acquisition.

Intrusion Detection Systems (IDS)،Auto Encoder،NIDS،Deep Learning،Network Traffic Analysis،Cyber Security