Generative Adversarial Networks for Addressing Data Privacy Issues

It is extremely important to preserve privacy of our citizens. One way to do it is to detect private information in the document and to indicate to owner of the documents that the documents contain privacy information. In order to develop machine learning algorithms to detect privacy data in the documents, the algorithms need to be trained with the existing documents that are annotated to point out private information. Access to those documents for training is limited since in many cases they are private as well. In addition, annotating large number of documents to indicate private information is extremely expensive and long process. Therefore, in this project we propose to develop algorithms for generating large number of documents that contain fictitious private information that will be used for training the algorithms. TO BE CONT'D

Intern: 
Rajitha Prabath
Faculty Supervisor: 
Miodrag Bolic
Province: 
Ontario
Partner: 
Partner University: 
Discipline: