Friday, May 22, 2020
A Federated Learning Approach for Mobile Packet Classification
Evita Bakopoulou, UCI
11:00am, Zoom (link to be sent via email) [Host: Scott]
In order to improve mobile data transparency, network-based approaches have been proposed to inspect packets generated by mobile devices and detect sensitive information, ad requests, tracking, etc. State-of-the-art approaches train classifiers based on features extracted from HTTP packets. So far, these classifiers have only been trained in a centralized way, where mobile users label and upload their packet logs to a central server, which then trains a global classifier and shares it with the mobile devices. Since these packet logs may contain sensitive information, we propose a federated learning approach. We discuss methodological challenges specific to this context and we evaluate our framework for two different tasks and using three real-world datasets.
Evita Bakopoulou received her B.Sc. and M.Sc. degrees in Computer Science from Athens University of Economics and Business, Greece, in 2014 and 2016, respectively. Currently, she is a fourth year Ph.D. student in the Networked Systems Program at UC Irvine, working with Prof. Athina Markopoulou. She has received the UCI Networked Systems Fellowship (2016-2018), H. Samueli Fellowship (2017-2018) and Broadcom Foundation Fellowship (2018-2019) for her graduate studies. Her research interests are primarily in the area of Machine Learning and Mobile Privacy. For more information, please visit her personal page (https://www.ics.uci.edu/~ebakopou/).