JMU // ISEG

Project Outline

In the project "DeepScan", together with the chair of data science of JMU, we developed a software to continuously and automatically check ERP systems for security-relevant irregular behavior. All while trying to limit the analysis to what is necessary to be legally compliant in a GDPR setup. The project makes use of artificial intelligence in form of neural networks to find common anomalies in enterprises. In order to train the AI model, the software first analyzes the data, identifies any potential instances of abuse, learns regular patterns, and then detects future missuse. We build our demonstrator on an active learning setup to reduce the needed human input and enrichted it with post-hoc Explainer to ease the labeling process. Following are the links to articles and code thatresulted from the Project.

Links:

• iNALU Paper: Frontiers in AI
• iNALU Code: github.com
• XAI Paper: link.springer.com
• XAI Code: github.com
• Software Architecture Paper (Beta): AIS Library
• PoC: https://jmu-iseg.github.io/anomalydetection/