Dr. Keke Chen, Northwestern Mutual Data Science Institute Associate Professor of Computer Science, and Yuechun (Ethan) Gu, a Marquette Ph.D. computer science student, were recently accepted to present at AAAI23, a top conference among AI and data science professionals. This year marks the 37th year of the conference, which will take place Feb. 7-14, in Washington, D.C.
Sponsored by the Association for the Advancement of Artificial Intelligence, AAAI brings together researchers and professionals in AI to highlight the latest research and trends in the industry while fostering intellectual interchange. The program features original research and practices with panel discussions and invited presenters exploring major social, philosophical, and economic issues influencing AI and its evolution around the world.
Chen and Gu will present their work on GAN-based domain inference attacks, introducing a method that would help protect corporate data against cyberattacks.
A generative adversarial network, or “GAN,” is a class of machine learning frameworks. In many instances, models are trained on historical company data, which often includes confidential data. Traditional model attacks looking to recreate a company’s training data are mostly carried out by attackers who have domain knowledge of the model itself.
Chen and Gu’s method explores the ability of substituting publicly available datasets within a GAN methodology in place of domain knowledge as a means to reverse-engineer the model and re-create the training data. Learn more and read their full paper online.