Doctoral Consortium
Community Detection of Anomalies in Large Scale Network using Deep Learning
Event Type
Doctoral Consortium
TimeTuesday, September 1511:45am - 12:30pm
LocationAyanna Howard Room
DescriptionAnomalies in network traffic is taking many different, and sometimes unrelated dimensions in these days of data-explosion phenomenon that produces “Big Data”. The line of demarcation between normal and abnormal data that are being transmitted and data that are at-rest is becoming thinner by the day because of various improvement in the levels of sophistication of cyber-security attacks.
The need for an equally sophisticated methods of cyber-security efforts necessitated recent developments in cyber-security solutions that combine both human and machine intelligence in addressing the explosive rate of cyber-attacks. The use of artificial intelligence (AI) in combating cyber-attacks cannot be over emphasized in this era.
This research is divided into two major parts. The first part aims at using deep learning anomaly detection technique as an artificial intelligence approach in strengthening the line of demarcation between normal and abnormal network traffic in a large time-series dataset. The second part plans to use network graph-based community detection approach to study the pattern of any existing or predicted relationship between the identified anomalies in the first part of the research.
Recent changes in cyber-threat’s landscape such as the increase in “living-off-the-land” attacks call for more sophisticated approach in providing cyber security solutions. This research proposal plans to implement a novel and robust solution that combines AI and cybersecurity to solve complex network security problems. The idea proposes the use of Long Short-Term Memory (LSTM), PageRank and Feature learning models to identify, group and predict anomalies in large scale real-world networks that contain millions or billions of nodes. The research also aims at providing insight into how the anomalies in the network interact.