Allen Riddell, an assistant professor of information science at the School of Informatics, Computing, and Engineering, is part of a team that has been awarded a grant from the National Science Foundation to study ways to protect whistleblowers from authorship attribution attacks.
The three-year grant, which is worth $481,000 and will be shared with researchers at Duquesne University, supports work developing methods to defend against techniques used to identify an anonymous author of documents such as letters, emails, memos, or social media posts through the analysis of writing style. An individual’s writing style—characterized by word choice and sentence structure—is often distinctive. Tell-tale “fingerprints” on documents can reveal the identity of whistleblowers and lessen the guarantees of anonymity and confidentiality offered by government-sponsored whistleblower programs.
“Protecting whistleblowers in areas where whistleblowing is an indispensable tool, such as the financial, transportation, and health care industries, is important,” Riddell said. “In these sectors, there's a clear desire by some whistleblowers to remain anonymous. If it’s trivial to unmask whistleblowers by studying their writing style, then people will, we think, be less likely to report illegal or harmful behavior.”
The research will use experiments to understand how to best help writers protect themselves from authorship attribution techniques and aims to develop software that would allow organizations, firms, and governments running whistleblower programs to better protect the identity of sources. Advances in quantitative and algorithmic approaches have improved the ability to unmask anonymous authors, but Riddell’s project seeks to validate a novel defensive technique and to create new tools for defending against authorship attribution attacks.
Riddell and his colleagues will study the effectiveness of known defenses against such attacks, create software to provide interactive feedback to sources who need to write prose which doesn’t reveal their identities, and develop a novel tool which supports automatic, non-interactive rewriting of text that preserves the semantic content of a document while removing identifying stylistic fingerprints.
The open-source software produced in the project can be disseminated to government agencies and other organizations to foster participation in whistleblower programs.
“Maintaining the anonymity of whistleblowers is critical to the oversight of our institutions,” said Ron Day, chair of the Information and Library Science program. “Allen’s project will play a key role in understanding how technology can serve as a protection against those who wish to peel back the veil of anonymity.”
To learn more about information science at SICE, visit our website.