Cyber threat detection using machine learning

Abstract: The project consists of a web-based application build in django (i.e. Python) that includes a dashboard for users to monitor cyberstalking. A machine learning classification algorithm (eg Support Vector Machine) can be trained to identify cyberstalking messages and then the classified messages can be imported into a database and is summarised on a dashboard. The dashboard displays timeseries data, topic models (for cyberstalking messages and non cyber bullying messages) and a summary of the affective dimensions found in the test messages (for cyberstalking messages and non cyber bullying messages). System must be scheduled to perform topic modeling and affective sentiment analysis. A moderation role that is able to mark classified messages as mis-classified. The project entitled “Cyber threat detection using machine learning” is developing an online cloud-based cyberstalking detection system. The system has to implement a cyberstalking detection system that integrates with popular social media services such as Facebook and Twitter. Researchers and IT professionals will have access to download, use, and test the opensource code license for this detection system that will be freely available on the Internet with simple installation instructions, and a highly user- friendly interactive dashboard that can be used by schools and parents as needed. This project will focus on the key natural language processing technologies that will allow traces of cyber-bullying to be captured and classified. The outcomes from this project will lead to free and easy-to-install software where the system picks up patterns in social media interactions that can be construed as potential cyber-bullying.

Keywords: Cyberstalking Detection, Machine Learning Classification, Support Vector Machine (SVM), Natural Language Processing (NLP), Django (Python Web Framework), Social Media Monitoring, Topic Modeling, Sentiment Analysis.