Machine Learning Approaches to Cyber Security, Markov Chain (MC)Model and Support Vector Machine (SVM) Approach

Bhaga, Khilari Satyawan (2016) Machine Learning Approaches to Cyber Security, Markov Chain (MC)Model and Support Vector Machine (SVM) Approach. Masters thesis, Indian Institute of Technology Hyderabad.

Text
EE14MTECH11034.pdf - Submitted Version
Restricted to Registered users only until 6 February 2020.
Download (1MB) | Request a copy

Abstract

In this thesis I presented machine learning application for cyber security. In particular anomalies are detected with Markov chain model technique and Support Vector Machine (SVM) method. Markov chain model form with normal or anomaly free data and considered as reference for anomaly detection. For anomaly detection ground truth (anomaly free) data is important which is lack in availability so I generated data in MATLAB and used to make MC (Markov Chain) model and for anomaly detection. I also used data generated by software SADIT downloaded from github. For SVM technique Kernel function used is ‘radial basis function’. SVM technique trained and tested with data generated by SADIT. Three dataset created of SADIT data to perform experiment using SVM method. MC model technique gives good performance for low noise level data but not gives good result for large noise level data. So MC model technique is robust for low noise level data. SVM technique not giving same results for all datasets i.e. for some datasets its performance is good and for some datasets its performance is not good. This is because features of data, so feature selection is important for SVM i.e. features for which SVM gives good performance that features should be selected.

[error in script]

IITH Creators: