Mandating the Forbidden: Balancing Privacy and Security in Production Abuse Systems

Monday, August 12, 2019 - 11:00 am11:30 am

Andy Schou, Google

Abstract: 

Production-scale abuse systems at major platforms are increasingly the subject of both regulatory mandates and privacy restrictions. Those systems often require algorithmic inspection of private user data, retention policy exceptions to longitudinally track major abuse campaigns, and large numbers of abuse raters with privileged access to production data. We discuss policy exceptions, corpus reduction, false positive minimization, and rater auditing and suggest privacy best practices for these critical but high-risk production systems.

Andy Schou, Google

Andreas Schou is a Staff Privacy Engineer in the Privacy and Data Protection Office at Google, focusing on Google’s abuse and machine learning systems. His work centers on preventing memorization and disclosure of user data by machine learning models and minimizing exceptional data handling in production abuse systems.

BibTeX
@inproceedings {238156,
author = {Andy Schou},
title = {Mandating the Forbidden: Balancing Privacy and Security in Production Abuse Systems},
booktitle = {2019 {USENIX} Conference on Privacy Engineering Practice and Respect ({PEPR} 19)},
year = {2019},
address = {Santa Clara, CA},
url = {https://www.usenix.org/node/238157},
publisher = {USENIX Association},
month = aug
}