Anomaly Detection in Dynamic Graphs

Project description

Dynamic graph anomaly detection focuses on identifying anomalous patterns and outliers in evolving graph structures. Unlike static graphs, dynamic graphs continuously change over time, with nodes and edges being added, removed, or modified. hence, this project aims to introduce an approach that can incorporate the time dimension into the feature generation process, such that we generate semantically meaningful temporal features for anomaly detection in graph-based data. The project involves creating a framework that can efficiently process and analyze dynamic graph data in real-time. This framework will utilize machine learning methods to identify nodes or subgraphs whose behaviour deviates significantly from the norm. By considering temporal aspects and changes in the graph's topology, the detection methods can pinpoint outliers that may indicate potential threats or unusual activities.

Further information

You may register interest in projects from different supervisors (not several projects with one supervisor). You must also contact each supervisor directly to discuss both the project details and your suitability to undertake the project.

Assumed knowledge

Pre-existing knowledge of graphs and graph-based data, programming in Python, and different machine learning/deep learning models.

Supervisors research focus

Anomaly Detection, Graph Data, Machine Learning and AI

Note: You need to register interest in projects from different supervisors (not a number of projects with the one supervisor).
You must also contact each supervisor directly to discuss both the project details and your suitability to undertake the project.