Company Filing History:
Years Active: 2025
Title: Hengrui Jia: Innovator in AI Privacy Protection
Introduction
Hengrui Jia is a notable inventor based in Toronto, Canada. He has made significant contributions to the field of artificial intelligence, particularly in the area of privacy protection for machine learning models. His innovative approach addresses critical concerns regarding data security and model integrity.
Latest Patents
Hengrui Jia holds a patent for the invention titled "Detection of privacy attacks on machine learning models." This patent describes a method where a plurality of queries are input into an AI model composed of multiple layers, including an input layer, an output layer, and at least one intermediate layer. During inference, each intermediate layer can output a plurality of activations. The invention intercepts activations from these intermediate layers for each query and determines if the distribution of these activations indicates attempts to manipulate the AI model. This is achieved through a distance-based similarity analysis between intercepted activations and reference activations. The results of this analysis are then provided to a consuming application or process, enhancing the security of AI systems.
Career Highlights
Hengrui Jia is currently associated with Hiddenlayer, Inc., where he continues to develop innovative solutions in AI. His work focuses on enhancing the robustness of machine learning models against privacy attacks, which is increasingly vital in today's data-driven world.
Collaborations
Hengrui has collaborated with talented individuals such as Ravikumar Balakrishnan and Zeliang Kan. Their combined expertise contributes to the advancement of privacy protection technologies in AI.
Conclusion
Hengrui Jia's contributions to the field of artificial intelligence, particularly through his patent on detecting privacy attacks, highlight his role as an innovator in this critical area. His work not only advances technology but also addresses essential privacy concerns in machine learning.