Creating deceptive Twitter bots
The increased usage and reliance of social media for news and information has been accompanied by attempts to influence opinions and conversations online.
This is often achieved on social media platforms like Twitter through the use of social bots: algorithms that exhibit human-like behaviour. In recent years, these bots have grown significantly in numbers and have become more sophisticated, which is why researchers have developed machine learning algorithms to detect them automatically.
This research considers how an adversary may discretely manipulate the behaviour of bot accounts to bypass bot detection. The purpose of this research is to highlight the vulnerabilities in existing detection algorithms so as to encourage their further development and enhance their security and trustworthiness. The performed evasion attacks are evaluated and defensive strategies are proposed in the hopes of improving the robustness of machine learning classifiers.
Securing our future
Electrical and Electronic Engineering