Review on Deep Learning Based IoT Intrusion Detection System

EasyChair Preprint no. 7408

Abstract

One of the goals of smart environments is to improve human life quality in terms of comfort and efficiency. The Internet of Things (IoT) standard has lately evolved into a smart environment technology. The key concerns in any real-world smart environment based on the IoT prototype are security and privacy. Security flaws in IoT-based systems could lead to security concerns infecting smart environment applications. As a result, there is a substantial need for IoT-specific intrusion detection systems (IDSs) to prevent IoT-related security threats that exploit only a handful of these security flaws. Traditional IDSs may not be a solution for IoT environments due to the restricted computation and storage capabilities of IoT devices, as well as the protocols employed. The increased awareness of vulnerabilities and associated attack pathways has an impact on a number of security goals. The major goal is to construct three abstraction levels of features, namely packet-based, unidirectional-based, and bidirectional-based features, that are determined. The evaluation process is carried out using a MQTT simulated dataset. The experimental findings indicated that ML models are capable of meeting the ID needs of MQTT-based networks.

Keyphrases: deep learning, Intrusion Detection System, KDD, log analysis

@Booklet{EasyChair:7408,
  author = {Rajeswari Somasundaram and P Karthikeyan},
  title = {Review on Deep Learning Based IoT Intrusion Detection System},
  howpublished = {EasyChair Preprint no. 7408},

  year = {EasyChair, 2022}}