Description
Blockchain is a distributed ledger, offering an innovative approach to establishing trust in a public and open environment. However, it bears many security attacks in sustaining data/status consistency in the process of data storing and exchanging, and the long-delay attack is one of the most challenging ones. Many researchers have adopted various methods in analyzing the blockchain system security by considering a upper-bounded and static network latency. In a realistic scenario, the network latency is always changing dynamically, resulting in the unreliability and inaccuracy of the existing analysis approaches. In this book, we propose a blockchain security analysis model based on partially observable Markov decision process (POMDP) against the long delay attack by capturing the dynamic network delay. In our model, an observation function about the network delay is learned and updated based on a clustering algorithm according to the real-time network status. With the support of the observation function, a POMDP model is constructed for attackers to maximize their expected rewards. To analyze the security of a blockchain system against the long delay attack, the utility of the attackers and normal miners with the same mining power is calculated and compared. The system is safe as the utility of the normal miners is no less than that of the attackers. Extensive experiments are conducted to show the effectiveness of the proposed analysis model, where Bitcoin system is evaluated and the safe ranges of the system network parameters are analyzed.