Benford’s Law can detect malicious social bots

Social bots are a growing presence and problem on social media. There is a burgeoning body of work on bot detection, often based in machine learning with a variety of sophisticated features. In this paper, we present a simple technique to detect bots: adherence with Benford’s Law. Benford’s Law states that, in naturally occurring systems, the frequency of numbers first digits is not evenly distributed. Numbers beginning with a 1 occur roughly 30 percent of the time, and are six times more common than numbers beginning with a 9. In earlier work, we established that Benford’s Law holds for social connections across online social networks. In this paper, we show that this principle can be used to detect bots because they violate the expected distribution. In three studies — an analysis of a large Russian botnet we discovered, and studies of purchased retweets on Twitter and purchased likes on Facebook — we show that bots’ social patterns consistently violate Benford’s Law while legitimate users follow it closely. Our results offer a computationally efficient new tool for bot detection. There are also broader implications for understanding fraudulent online behavior. Benford’s Law is present in many aspects of online social interactions, and looking for violations of the distribution holds promise for a range of new applications.