Speaker: Sadia Afroz (Berkeley EECS Department)
Title: Anonymity in the big data era
Host: Joan Feigenbaum
Abstract: With the rise of the social web, sharing data has become an integral part of the internet. There has been an explosion in user-generated content to express opinions, to coordinate protests against repressive regimes, to whistle blow, to commit fraud, and to disclose everyday information. At the same time, this information is being misused without the users’ consent by advertisement campaigns for targeted ads, by cybercriminals for targeted spamming and are inhibited by many state level censors.
My research aims to improve the anonymity and privacy of internet users. In this talk I will focus on the privacy threat with writing style. Current anonymity systems focus strongly on location-based privacy but do not address many avenues for the leakage, especially identification through the content of data. I will demonstrate how we linked different accounts of real-world cybercriminals by analyzing only their writing style, even though the cybercriminals used different usernames and email addresses to hide their identities. I will also show how this attack can be evaded by changing writing style and discuss our tool “Anonymouth” that can help anonymizing writing style.
Bio: Sadia Afroz, PhD, is a postdoctoral researcher at the University of California, Berkeley. She graduated from Drexel University, Philadelphia in 2014. Her research focuses on the adversarial aspect of machine learning and the privacy of writing style. Her work on adversarial authorship attribution was selected as a runner-up for the 2014 ACM SIGSAC dissertation award, the 2013 Privacy Enhancing Technology (PET) award and the best student paper award at the Privacy Enhancing Technology Symposium (PETS) 2012. More about her research can be found: https://www.eecs.berkeley.edu/~sa499/