Network security mechanisms and implementations for the next generation reliable fast data transfer protocol - UDT

Publication Type:
Issue Date:
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail01Front.pdf136.95 kB
Adobe PDF
Thumbnail02Whole.pdf1.52 MB
Adobe PDF
TCP protocol variants (such as FAST, BiC, XCP, Scalable and High Speed) have demonstrated improved performance in simulation and in several limited network experiments. However, practical use of these protocols is still very limited because of implementation and installation difficulties. Users who require to transfer bulk data (e.g., in Cloud/GRID computing) usually turn to application level solutions where these variants do not fair well. Among protocols considered in the application level are User Datagram Protocol (UDP)-based protocols, such as UDT (UDP-based Data Transport Protocol). UDT is one of the most recently developed new transport protocols with congestion control algorithms. It was developed to support next generation high-speed networks, including wide area optical networks. It is considered a state-of-the-art protocol, addressing infrastructure requirements for transmitting data in high-speed networks. Its development, however, creates new vulnerabilities because like many other protocols, it relies solely on the existing security mechanisms for current protocols such as the Transmission Control Protocol (TCP) and UDP. Certainly, both UDT and the decades-old TCP/UDP lack a well-thought-out security architecture that addresses problems in today’s networks. In this dissertation, we focus on investigating UDT security issues and offer important contributions to the field of network security. The choice of UDT is significant for several reasons: UDT as a newly designed next generation protocol is considered one of the most promising and fastest protocols ever created that operates on top of the UDP protocol. It is a reliable UDP-based application-level data-transport protocol intended for distributing data intensive applications over wide area high-speed networks. It can transfer data in a highly configurable framework and can accommodate various congestion control algorithms. Its proven success at transferring terabytes of data gathered from outer space across long distances is a testament to its significant commercial promise. In this work, our objective is to examine a range of security methods used on existing mature protocols such as TCP and UDP and evaluate their viability for UDT. We highlight the security limitations of UDT and determine the threshold of feasible security schemes within the constraints under which UDT was designed and developed. Subsequently, we provide ways of securing applications and traffic using UDT protocol, and offer recommendations for securing UDT. We create security mechanisms tailored for UDT and propose a new security architecture that can assist network designers, security investigators, and users who want to incorporate security when implementing UDT across wide area networks. We then conduct practical experiments on UDT using our security mechanisms and explore the use of other existing security mechanisms used on TCP/UDP for UDT. To analyse the security mechanisms, we carry out a formal proof of correctness to assist us in determining their applicability by using Protocol Composition Logic (PCL). This approach is modular, comprising a separate proof of each protocol section and providing insight into the network environment in which each section can be reliably employed. Moreover, the proof holds for a variety of failure recovery strategies and other implementation and configuration options. We derive our technique from the PCL on TLS and Kerberos in the literature. We maintain, however, the novelty of our work for UDT particularly our newly developed mechanisms such as UDT-AO, UDT-DTLS, UDT-Kerberos (GSS-API) specifically for UDT, which all now form our proposed UDT security architecture. We further analyse this architecture using rewrite systems and automata. We outline and use symbolic analysis approach to effectively verify our proposed architecture. This approach allows dataflow replication in the implementation of selected mechanisms that are integrated into the proposed architecture. We consider this approach effective by utilising the properties of the rewrite systems to represent specific flows within the architecture to present a theoretical and reliable method to perform the analysis. We introduce abstract representations of the components that compose the architecture and conduct our investigation, through structural, semantics and query analyses. The result of this work, which is first in the literature, is a more robust theoretical and practical representation of a security architecture of UDT, viable to work with other high speed network protocols.
Please use this identifier to cite or link to this item: