Information digitalization brings huge opportunities but also makes information vulnerable. A major challenge for computer scientists is to build a trustworthy digital world. Satoshi Nakamoto, whoever that is, proposed a genius idea: to store information, in the form of a chain of blocks, at all nodes of a distributed computing system. In Nakamoto’s protocol, called the Blockchain protocol, a block is a sequence of transactions, each representing a transfer of assets – Bitcoins – from one user to another. Nakamoto’s approach can be straightforwardly generalized to any fungible asset, leading to a tokenized peer-to-peer system. Each token can represent any type of material or digital good and is associated with a particular user, the owner of that token. Even if participants do not trust each other, they can reliably trust that the content of the blockchain is consistent among them, provided that the majority is honest. This trust property is invaluable towards implementing a wide range of decentralized applications, such as payments, crowdfunding, or insurance contracts. Although Nakamoto’s original protocol has spurred significant innovation and financial interest in the last decade, its computational cost has been prohibitive. The protocol has been estimated to consume as much electricity as a modern, industrialized, country. Transaction latencies have been prohibitively high and the system has sustained a very low throughput. Hundreds of alternative solutions for implementing decentralized trust have been proposed since the original Blockchain protocol: each alternative seeks to reduce energy consumption, obtain lower latency, or improve throughput. All proposed alternatives, however, sacrifice either trustworthiness or efficiency.
In fact, it is not surprising that the original blockchain protocol and its alternatives have been so computation expensive. All these solutions seeked to solve a notoriously difficult problem: consensus. Essentially, the set of nodes in the network have to agree on the same value, e.g. the position of a block in the chain, despite the possibility of malicious behavior of some of the nodes, or network delays. The consensus problem has been the most studied problem in distributed computing because of its notorious difficulty, and many impossibility and lower bound results were established. In practical terms, these results translate into inherent trade-offs between trust and efficiency. In the context of our ERC AOC (Adversary-Oriented Computing) project (advanced grant), we revisited the issue of implementing a trustworthy payment system, i.e. the problem solved in Nakamoto’s paper by « his » Blockchain protocol. We made a crucial discovery: solving the difficult consensus problem is not needed for building a trustworthy payment system. It is enough to solve a problem called secure causal broadcast to implement digital trust. This problem is significantly simpler than consensus and is not subject to classical impossibility results and lower bounds.
In the context of this project, we solved that problem and devised a scalable protocol for asset transfers that does not consume more energy than the sending of a few messages on the Internet. Our protocol, named AT2, is generic, and comes in several flavours depending on the target environment.
