Date of Award
8-2017
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical and Computer Engineering (Holcomb Dept. of)
Committee Member
Dr. Richard Brooks, Committee Chair
Committee Member
Dr. Pierluigi Pisu
Committee Member
Dr. Yingjie Lao
Committee Member
Dr. Yongqiang Wang
Committee Member
Dr. Robert Lund
Abstract
Botnets have been problematic for over a decade. They are used to launch malicious activities including DDoS (Distributed-Denial-of-Service), spamming, identity theft, unauthorized bitcoin mining and malware distribution. A recent nation-wide DDoS attacks caused by the Mirai botnet on 10/21/2016 involving 10s of millions of IP addresses took down Twitter, Spotify, Reddit, The New York Times, Pinterest, PayPal and other major websites. In response to take-down campaigns by security personnel, botmasters have developed technologies to evade detection. The most widely used evasion technique is DNS fast-flux, where the botmaster frequently changes the mapping between domain names and IP addresses of the C&C server so that it will be too late or too costly to trace the C&C server locations. Domain names generated with Domain Generation Algorithms (DGAs) are used as the 'rendezvous' points between botmasters and bots. This work focuses on how to apply botnet technologies (fast-flux and DGA) to counteract network traffic analysis, therefore protecting user privacy. A better understanding of botnet technologies also helps us be pro-active in defending against botnets. First, we proposed two new DGAs using hidden Markov models (HMMs) and Probabilistic Context-Free Grammars (PCFGs) which can evade current detection methods and systems. Also, we developed two HMM-based DGA detection methods that can detect the botnet DGA-generated domain names with/without training sets. This helps security personnel understand the botnet phenomenon and develop pro-active tools to detect botnets. Second, we developed a distributed proxy system using fast-flux to evade national censorship and surveillance. The goal is to help journalists, human right advocates and NGOs in West Africa to have a secure and free Internet. Then we developed a covert data transport protocol to transform arbitrary message into real DNS traffic. We encode the message into benign-looking domain names generated by an HMM, which represents the statistical features of legitimate domain names. This can be used to evade Deep Packet Inspection (DPI) and protect user privacy in a two-way communication. Both applications serve as examples of applying botnet technologies to legitimate use. Finally, we proposed a new protocol obfuscation technique by transforming arbitrary network protocol into another (Network Time Protocol and a video game protocol of Minecraft as examples) in terms of packet syntax and side-channel features (inter-packet delay and packet size). This research uses botnet technologies to help normal users have secure and private communications over the Internet. From our botnet research, we conclude that network traffic is a malleable and artificial construct. Although existing patterns are easy to detect and characterize, they are also subject to modification and mimicry. This means that we can construct transducers to make any communication pattern look like any other communication pattern. This is neither bad nor good for security. It is a fact that we need to accept and use as best we can.
Recommended Citation
Fu, Yu, "Using Botnet Technologies to Counteract Network Traffic Analysis" (2017). All Dissertations. 2015.
https://tigerprints.clemson.edu/all_dissertations/2015