Abstract
Distributed Transient Networks (DTNs) are the type of network that is inherently decentralised by nature, and mainly consists of equally-privileged client users who are able to join in and leave the network at any time. Users are evenly assigned with the same computation tasks, such as routing messages and manage topological information, without bias. Typical examples of DTN include Peer-to-Peer (P2P) Networks, Delay Tolerant Networks, Vehicular Ad Hoc Networks (VANETs) and many others. Depending on how they use network infrastructure, we divide DTN applications into three categories, which are Infrastructure-based DTN, Infrastructure-less DTN, and Hybrid DTN. However, there are some peer communication latency issues in any type of DTN. In this thesis, our overall aim is to improve the reliability of DTNs, which is degraded by long communication latency, from the perspectives of augmenting the performance of fundamental routing and the security of DTN devices.
This thesis firstly enhances the routing performance of high latency DTNs by proposing a novel probabilistic routing protocol for DTNs, namely, FGAR. It characterises the contact patterns among peers in a fine-grained memory-efficient manner, and thus gains the ability to produce a flexible future contact prediction, based on the history within specified timeframes. In addition, a dynamic message replication scheme is also built on the above probabilistic estimation as part of the FGAR protocol, which intelligently eliminates much of the unnecessary message copy generation, without introducing too much computational complexity. Through the trace-driven simulations, FGAR reduces network overhead up to 15 times, and improves the message delivery ratio more than 20% compared to several popular existing routing protocols of DTNs.
Blockchain technology, the distributed ledger and the fundamental security scheme of bitcoin, provides a secure decentralised environment among untrusted users by maintaining an immutable global consensus contributed to by any user that chooses to participate. In this thesis, we investigate the impact of network latency on the safety of bitcoin blockchain, which is primarily evaluated in terms of the time spent on block convergence, and the possibility of violating the six confirmations convention. The conducted simulation shows that with a longer communication latency, blockchain suffers an exponential delay on the convergence speed and more revoked blocks. In addition, peers with lower latency gain unfair advantages in competing for the PoW reward, as they have a higher chance of obtaining the reward by contributing to the blockchain compared to the group of users with high-latency connections to the network.
Avalanche is a lightweight decentralised consensus protocol that focuses on guiding the entire network towards a global agreement for each conflicting event. We explore the impact of network latency on the safety of Avalanche within three different network environments. Firstly, nodes have various degrees of communication latency. Secondly, the network is divided into two communities, with a precisely controlled inter-communication frequency between them. Lastly, the communication among nodes is achieved via a pure hop-by-hop manner. The simulated results illustrate that Avalanche has an outstanding capability for operating correctly even with high communication latency. However, Avalanche becomes more fragile when opinions are passed from one network partition to another, so network topology may be an important consideration. Additionally, we also propose a general mathematical relation between the security strength of Avalanche with the number of transmission hops taken to deliver a message. Our mathematical analysis reveals that the security strength of Avalanche is exponentially weakened with more intermediate nodes involved in relaying the consensus queries. Accordingly, our mathematical analysis is well supported by the reported statistics from the simulations. Our key findings indicate that communication latency, nodes' reachability and multi-hop communication are the major factors that significantly impact the security strength of decentralised consensus. The findings provide us with an important direction to design a new decentralised consensus for DTNs in future work.