Confirmed or Dropped? Reliability Analysis of Transactions in PoW Blockchains

Blockchains based on Proof-of-Work (PoW) have introduced a new paradigm for distributed ledgers on the Web. In these systems, transactions compete to obtain a position inside the new blocks by offering a fee to be confirmed before others. A finite amount of memory is devoted to store unconfirmed transactions, called Mempool. When new transactions arrive and the Mempool is full, silent droppings of the cheapest unconfirmed transactions occur, without any notification to the owners. This challenge becomes particularly pressing as users have the freedom to append various types of data to the blockchain, including large media files, leading to swift Mempool depletion. In this paper, we study the reliability of PoW blockchains from a user perspective. We provide a numerical model to answer the question: What is the probability of confirmation for a transaction offering a fee $f$ when the system is in a certain state? Our model allows blockchain-based applications to analyse the tradeoff between running costs and reliability, i.e., fees offered for the transactions and probability that the transactions will be eventually confirmed. The proposed method is proactive and does not require historical data on dropped transactions that, in fact, are not logged anywhere in the blockchain. This paper presents significant contributions, summarized as follows: (i) the introduction of a stochastic model and its efficient solution for analyzing dropping probability in blockchain systems; (ii) validation of the model through real traces extracted from the Bitcoin blockchain.