Covert Botnet Command and Control Using Twitter

Abstract

Botnets are one of the primary threats in computer security today. They are used for launching denial of service attacks, sending spam and phishing emails, and collecting private information. However, every botnet requires coordination. In order to initiate an attack, a botmaster must communicate to all of the bots in the network. In this paper, we present a steganographic system that demonstrates the feasibility of the social networking website Twitter as a botnet command and control center that an attacker could use to reliably communicate messages to a botnet with low latency and nearly perfect rate of transmission. Our system generates plausible cover messages based on a required tweet length determined by an encoding map that has been constructed based on the structure of the secret messages. The system considers both the input symbol frequencies (e.g. English letter frequencies) as well as the tweet length posting frequencies for constructing the encoding maps. A technique for automatically generating Twitter account names based on Markov chains is also presented so that the bots can connect to new accounts if the existing botmaster account is unavailable. All the experiments were performed using the 7.3M actual tweets from 3.7K verified accounts collected by the tweet parser developed by us. We have evaluated the efficacy of the system using Emulab and usability of the system through Amazon's Mechanical Turk with promising results. An analysis of the steganographic security of the proposed system has also been provided. By demonstrating how a botmaster might perform such communication using online social networks, our work provides the basis to detect and prevent emerging botnet activities.