Brennan is a Ph.D. candidate in the computer science department and is working in the Privacy, Security and Automation Lab on stylometry and its applications in artificial intelligence.
A set of new software programs developed by computer scientists at Drexel could soon help protect the speech of the disenfranchised and defend the voice of the whistleblower—all by confirming or contorting one’s writing style.
JStylo and Anonymouth are a set of competing programs designed, respectively, to confirm the authorship of a disputed text and help authors to remain anonymous. The open-source software was developed in Drexel’s Privacy, Security and Automation Lab.
“JStylo and Anonymouth are intended to fill holes that we see in both the research and privacy communities,” said Michael Brennan, a lead developer of the project and fifth-year doctoral student at Drexel. “JStylo allows for more effective stylometry research and Anonymouth enables people to maintain their anonymity when publishing sensitive writing.”
JStylo, the program intended to identify authors, uses a rigorous set of filters to sift out patterns in the text. Starting with broad categories such as sentence length and lines per paragraph down to such characteristics as word choice and frequency of certain letter combinations, the program generates an author profile, which is then compared to a baseline writing sample from the suspected author. With a writing sample of about 6,500 words as a comparison, the software can select an author from a pool of 40 candidates with 80-85 percent accuracy, according to Brennan. The tool is even more accurate as the number of possible authors is decreased.
Conversely, Anonymouth is designed to help cloak an author’s unique writing characteristics to a point where the text could not be traced back to them using authorship recognition software—such as JStylo. Anonymouth does not encode the writing, but it goes through a similar set of analyses to its counterpart and suggests changes that the author could make in order to mask his or her writing style.
“When people want to speak anonymously, whether it be for reporting on human rights issues or whistleblowing or simply voicing unpopular opinions, they need to know how to be safe and whether stylometry may reveal their identity,” said Dr. Rachel Greenstadt, the director of Drexel’s Privacy, Security and Automation Lab.
The products’ recent launch at the Chaos Communication Congress marked the first release of an open-code software of this kind. Brennan and Greenstadt are projecting a beta release this spring and an accompanying research paper, titled “Detecting Hoaxes, Frauds, and Deception in Writing Style Online,” was recently accepted to the IEEE Symposium on Security and Privacy, a fairly prestigious conference for this field, Brennan said.