Research And Design Of Dynamic Byzantine Fault Tolerance Protocol

Recently,with the astonishing tide of digital currency led by Bitcoin,people attach great importance to the robustness of "blockchain",a distributed system.How to solve the consensus problem with the participation of adversaries,and design a practical BFT(Byzantine Fault Tolerance)protocol has become a hot research topic.In this thesis,we analyze and summarize the main researches of BFT protocol,and study in depth two representative protocols,PBFT and HoneyBadgerBFT.During our research,we find the following problems.Firstly,these protocols do not allow nodes to join or exit dynamically.They all rely on a completely closed system,where any node wants to join or exit,the only way is to stop the whole system and modify system parameters.Unfortunately,frequent downtime updates are obviously intolerable for high-traffic systems in use.In addition,these protocols all lack a clear mechanism to judge the activity of consensus nodes.The truth that nodes are whether active or not has a great impact on the system security.Similar to malicious nodes,inactive ones also seriously threaten the stability of the whole system.Because of this,malicious or inactive nodes should be detected and handled in time.Worse still,these protocols do not take any penalties to deal with malicious nodes.The destructive power of these nodes will increase constantly and control the honest ones.Eventually,these nodes will cause system crash.Problems above reduce the practicability of BFT protocol and need to be solved urgently.This thesis considers the advantages and disadvantages of existing protocols and improves them,which comes to following conclusions:1.Based on PBFT protocol,we design a novel dynamic BFT protocol,named "Dynamic PBFT Protocol".This protocol retains the normal process of PBFT to obtain good fault tolerance and security.Moreover,the protocol innovatively proposes the Network Dynamic Consensus protocol to update the node information,which is useful for nodes to dynamically join or exit.In addition,the protocol designs a perfect mechanism to quantify the activity of nodes.By introducing the concept of "Participation",we can quantitatively analyze the contribution of nodes.As a result,malicious or inactive nodes can be quickly discovered.Finally,the protocol also sets some penalties for these nodes.Upon finding malicious or inactive nodes,the system will not only remove them,but also report them in a blacklist and do not allow them to join again.Through the timely detection and processing of malicious nodes,our protocol improves the stability and security of the system greatly.2.We also propose a practical dynamic asynchronous BFT protocol.Besides designing the Dynamic PBFT protocol for weakly synchronous environments,this thesis also improves the HoneyBadgerBFT protocol and proposes a dynamic BFT protocol for fully asynchronous networks.Firstly,our protocol inherits the excellent asynchronism and fault tolerance of the original protocol.Additionally,by introducing a CA node,our protocol also achieves great dynamic property.This property not only meets the demands of node updating,but also facilitates the punishment and removing of malicious nodes.Similarly,dynamic asynchronous BFT protocol designs a detailed node inspecting mechanism,which is useful to discover malicious or inactive nodes and punish them.With the above measures,our protocol not only has great fault tolerance and dynamic property,but enhances the security and practicability of the system as well.