KAIROS RESEARCH LLC — Department of Defense STTR Phase I: N20A-T017

Description

The Kairos Research Team, together with our partners at the University of Arkansas at Little Rock (UALR), proposes to develop an intuitive web-based Twitter Follower / Friend Assessment (TWIFFA) tool and associated app that can identify bots and bot-assisted accounts within a user’s network of followers and friends. TWIFFA will enable a user to input their Twitter credentials, scan followers and friends for signs of bots or bot-assisted accounts, and assess the level of threat associated with flagged accounts. Innovative features of TWIFFA include: (1) explainable, multi-class bot detection algorithms that utilize a tiered approach to achieve high accuracy without sacrificing scalability; (2) frequent retraining of detection algorithms via user feedback to ensure robustness against constantly evolving bot behaviors; (3) multi-level threat assessment combining network-level focal structure analysis (FSA) with individual-level measures of a node’s “power” and role within the network; (4) operational response capabilities such as the ability to monitor, block and unfollow bot and bot-assisted accounts in batches, as well as the ability to label and track “watchlists” of suspected botnets involved in coordinated information campaigns.