Overview / Usage

Today, network and system security are of paramount importance in the digital communication environment. On par with the developments in technology, many threats to information security have emerged that can impact sensitive transactions: intruders can easily cause all kinds of breaches and crash networks, launch denial of service attacks, inject malware, and so on. To avoid such breaches, security administrators badly need the means to detect intruders and prevent them from entering the network. Nowadays, new threats and associated solutions are emerging together.
This session showcases a hybrid intrusion detection system that leverages the benefits of machine learning techniques to build a system that detects intrusion and alerts network administrators. This system can be extended from intrusion to breach detection as well.

The developed system analyzes and predicts user behavior, which in turn classifies as an anomaly or normal behavior. Other systems use just one machine learning algorithm to solve the problem, while this hybrid intrusion detection system uses a combination of algorithms for classification, achieving greater accuracy.

I have been working on this project for nearly 1.5 years and presented in Intel IoT Global Devfest 2018 as well.
The system is completely developed using Intel distribution of Python, Intel optimized TensorFlow and caffe. The model was trained on DevCloud.

Methodology / Approach

The developed system analyzes and predicts user behavior, which in turn classifies as an anomaly or normal behavior. Other systems use just one machine learning algorithm to solve the problem, while this hybrid intrusion detection system uses a combination of algorithms for classification, achieving greater accuracy.

Technologies Used

Python, TensorFlow, Caffe, Numpy, Scipy, Matplotlib