DETECTING ANOMALIES AND YAGADA

CSIT carries out extensive research into anomaly detection in the fields of video analysis, security monitoring, trajectory analysis and transaction monitoring.

This webinar, delivered on Tuesday 12th August 2014, features Dr Michael Loughlin (Senior Engineering Manager) and Dr Bhargav Mitra (Senior Engineer) introducing concepts associated with anomaly detection, the current state of the art as well as introducing the YAGADA anomaly detection algorithm developed at CSIT by Dr Michael Davis during the course of his PhD study here. Read his "Detecting anomalies in graphs with numeric labels" paper here.

YAGADA

Yagada (Yet Another Graph-based Anomaly Detection Algorithm) is an algorithm for detecting anomalies in labelled graphs and complex networks, such as computer networks, communication networks, social networks, or even networks of biological or ecological processes.

Yagada can be applied to any data which is represented as a set of entities (nodes) and the relationships between them (edges); for example, in a social network, nodes represent people and edges can represent relationships or interactions between people. Yagada finds the patterns which best describe the data and then uses density-based outlier detection combined with information theory to find anomalous regions in the network.

As nodes and edges can be labelled with attributes (e.g. job role, age, no. of interactions), Yagada is capable of processing complex, heterogeneous data with categorical, numeric and relational elements.