Abstract

The number of papers submitted to scientific conferences is steadily rising in many disciplines. To handle this growth, systems for automatic paper-reviewer assignments are increasingly used during the reviewing process. These systems employ statistical topic models to characterize the papers' content and automate their assignment to reviewers. In this talk, we invesitgate the security of this automation and introduce a new attack that modifies a given paper so that it selects its own reviewers. Our attack is based on a novel optimization strategy that fools the topic model with unobtrusive changes to the paper's content. In an empirical evaluation with a (simulated) conference, our attack successfully selects and removes reviewers, while the tampered papers remain plausible and often indistinguishable from innocuous submissions.