Intrusion detection is a big part of network security. Our project aims to solve this problem by detecting intrusion attacks as they happen using machine learning.
We are using the probabilistic model to detect an intruder as an outlier from our probability distribution curve.
We used kdd99 network dataset to check the various features helping in detection of intrusions. Using weight of these features, we sort out the relevant and more impactful features to improve the accuracy.
Our project detects intrusions without any signatures or predefined attacks.
This intrusion detection is based on outlier identification i.e. we are able to detect not only the known attacks but also the unknown attacks which may come in the future.