Unifying deep learning and knowledge representation for cybersecurity applications.

Companies are increasingly generating more data. There’s increasing value in sharing and analysing this data. However, it often contains personal and sensitive information and it’s very important to protect customer privacy by not sharing this information with third parties. This leads to a catch 22 situation. Companies have data they want to share but they can’t because of what’s in the data. Automated tools can be used to analyse the data and redact the sensitive information. However, they can’t currently be relied on because they’re not accurate enough. This is because accurate classification often depends on understanding context, which is something that computer systems struggle with. This project aims to design a new type of automated classification system that fuses best in class topological, contextual data analysis and natural language processing algorithms. This system will be able to understand both the structural and contextual information that existing automated systems can’t handle. This will allow personal and sensitive information to be identified much more accurately, which in turn will allow companies to share and extract value from their data.