Malware Analysis using Machine Learning

The objective of this project is to perform malware analysis on a given dataset to implement various basic classification-based machine learning algorithms. We make use of two datasets, one which is self-generated and another obtained from the internet. The classification algorithms used are Gaussian Naive Bayes, RandomForestClassifier, DecisionTreeClassifier & Linear SVC.

The self-generated dataset was created by:

• running windows PE files through the cuckoo sandbox to obtain benign file data
• obtaining malign hashes from virusshare.com and running them through an online malware analyzer (virustotal.com) to obtain malignant file data

The online dataset used was: https://www.kaggle.com/amauricio/pe-files-malwares

The accuracy obtained using various algorithms on the self-generated dataset:

• using Gaussian Naive Bayes: 50.5%
• using RandomForestClassifier: 96.96%
• using DecisionTreeClassifier: 100%
• using Linear SVC: 100%

The accuracy obtained using various algorithms on the online dataset:

• using Gaussian Naive Bayes: 32.24%
• using RandomForestClassifier: 98.44%
• using DecisionTreeClassifier: 41.01%
• using Linear SVC: 96.06%

• Include instructions for using Cuckoo Sandbox & storing data obtained in csv form.