Tracking and Mitigation of Malicious Remote Control Networks

Attacks against end-users are one of the negative side effects of today’s networks. The goal of the attacker is to compromise the victim’s machine and obtain control over it. This machine is then used to carry out denial-of-service attacks, to send out spam mails, or for other nefarious purposes. From an attacker’s point of view, this kind of attack is even more efficient if she manages to compromise a large number of machines in parallel. In order to control all these machines, she establishes a "malicious remote control network", i.e., a mechanism that enables an attacker the control over a large number of compromised machines for illicit activities. The most common type of these networks observed so far are so called "botnets". Since these networks are one of the main factors behind current abuses on the Internet, we need to find novel approaches to stop them in an automated and efficient way. In this thesis we focus on this open problem and propose a general root cause methodology to stop malicious remote control networks. The basic idea of our method consists of three steps. In the first step, we use "honeypots" to collect information. A honeypot is an information system resource whose value lies in unauthorized or illicit use of that resource. This technique enables us to study current attacks on the Internet and we can for example capture samples of autonomous spreading malware ("malicious software") in an automated way. We analyze the collected data to extract information about the remote control mechanism in an automated fashion. For example, we utilize an automated binary analysis tool to find the Command & Control (C&C) server that is used to send commands to the infected machines. In the second step, we use the extracted information to infiltrate the malicious remote control networks. This can for example be implemented by impersonating as a bot and infiltrating the remote control channel. Finally, in the third step we use the information collected during the infiltration phase to mitigate the network, e.g., by shutting down the remote control channel such that the attacker cannot send commands to the compromised machines. In this thesis we show the practical feasibility of this method. We examine different kinds of malicious remote control networks and discuss how we can track all of them in an automated way. As a first example, we study botnets that use a central C&C server: We illustrate how the three steps can be implemented in practice and present empirical measurement results obtained on the Internet. Second, we investigate botnets that use a peer-to-peer based communication channel. Mitigating these botnets is harder since no central C&C server exists which could be taken offline. Nevertheless, our methodology can also be applied to this kind of networks and we present empirical measurement results substantiating our method. Third, we study fast-flux service networks. The idea behind these networks is that the attacker does not directly abuse the compromised machines, but uses them to establish a proxy network on top of these machines to enable a robust hosting infrastructure. Our method can be applied to this novel kind of malicious remote control networks and we present empirical results supporting this claim. We anticipate that the methodology proposed in this thesis can also be used to track and mitigate other kinds of malicious remote control networks