Difficulty Prediction for Proof-of-Work Based Blockchains

Abstract

The fast growing of blockchain technologies has inspired an exceptionally broad set of new social applications due to its special features. The most two popular blockchains (Bitcoin and Ethereum) are Proof-of-Work (PoW) based. However, the randomness of the block produce time (BPT) greatly affects the performance, user experience, and even the security of a blockchain. Mining difficulty controls the stability of the BPT. However, the current difficulty control algorithm in Ethereum has two problems. First, the difficulty control algorithm cannot track the network hashrate fast enough. Thus, the BPT has a large variance and cannot converge to the target BPT on time. The second problem is that the target BPT cannot be flexibly set to any given value by adjusting the difficulty. To mitigate these two problems, we propose a linear predictor based difficulty control algorithm. Based on the relationship among the hashrate, the difficulty, and the BPT, the prediction based difficulty control algorithm has much better stability and flexibility on BPT. It is also shown that the proposed prediction based algorithm outperforms the existing Ethereum one in the real situation.