Software Quality and Security
Principal Investigator: Rudolf Ferenc
The proposed research project addresses four goals within the area of software quality assurance: a deep learning based bug prediction, security vulnerability prediction, code duplication filtering and test running based fault localization. All the topics targets the special software components of typical smart systems.
With applying static analysis and repository mining tools, we build databases both for software bugs and vulnerabilities and a set of static metrics of their source codes. Using machine learning (for example deep learning), we build prediction models that will be able to forecast potentially error-prone and vulnerable code components. We also propose a deep learning based model and algorithm to filter code clone instances detected by our existing tool.
To enhance state-of-the-art fault localization approaches, we propose a novel algorithm that uses call-chains collected by running unit tests. For this, we also build a manually validated set of real software defects.
Gergő Balogh, Árpád Beszédes, Tamás Gergely, Péter Gyimesi, Péter Hegedűs, Ferenc Horváth, Judit Jász, Edit Pengő, István Siket, Zoltán Tóth, Béla Vancsics