DLA-ABIDS:Deep Learning Approach for Anomaly Based Intrusion Detection System
Imen Ben Ahmed, Farah Barika Ktata, Khalil Ben Kalboussi · 2023
Nowadays, with the proliferation of the number of IoT devices, management and security of data are becoming crucial tasks. Intrusion detection systems (IDS) monitor network traffic for any unusual activity and send out alerts when it detects anomalies. The often-used intrusion detection systems are built on a variety of machine learning algorithms that allow the automation of detection on a scale that has never been achieved before. However, due to the massive size of traffic data and the nature of zero-day attacks, it is difficult to discover potential threats exploiting security vulnerabilities, which makes the detection process complicated. As a result, traditional IDS produce a high rate of false positive alerts. The suggested approach for anomaly intrusion detection problems, including zero-day attacks, uses a combination of classification and clustering machine learning techniques, such as a hybrid CNN-LSTM architecture for binary classification for real-time packet traffic clustering. The performed Experiments utilizes the Ton IoT 2019 Data set CSV files that present the IOT Network Traffic. The results demonstrated the efficiency of the proposed approach, which provides a convenient way to evaluate risks and vulnerabilities from CSV files, enabling adoption of real time network traffic with a low number of the false positive rate.