This study analyzes two covert Chinese bot networks, employing tweet-based and account-based methods to find detection evasion tactics. We reveal the use of message artifacts that disguise spam, engagement strategies that mimic human interaction, and behavioral patterns suggesting algorithmic control. We uncover bot maintenance practices and algorithmic account naming conventions. These insights demonstrate the evolving strategies of inauthentic digital personas, enhance our understanding of online disinformation campaigns, and inform the development of digital manipulation countermeasures. Comparing campaigns in 2021 and 2023, we discover that the techniques used by state-sponsored actors shifted from text-based to image-based techniques, indicating the increased sophistication of these actors to evade the detection algorithms of the social media platform. This work provides insight into the tactics of covert bot networks and discusses possible advancements in detection techniques.

PDF