Automating an Infosec Question Answering Function

Tugboat has 1000s of questions and responses submitted by clients. Users upload questions which can be answered using the “Auto Answer” functionality. The “Auto Answer” technology leverages weighted search vectors to search answer databases and prioritize which answers to recommend for a given question.

The objective of the project is to leverage this dataset by training a machine learning model to automate and improve the “Auto Answer” functionality described above. The current approach uses ideas from the information retrieval domain serving as the baseline. Improving this benchmark is the primary goal. The task of the model is to, given a question, recommend an ordered list of similar questions with known responses from the available databases, along with confidence or probability levels of the answers.

Faculty Supervisor:

Nathan Sturtevant

Student:

Paritosh Goyal

Partner:

Tugboat Logic Inc

Discipline:

Computer science

Sector:

Other

University:

University of Alberta