Multi-Layer Hash Anchoring method with waterfall consensus for fact-oriented blockchain

The paper proposes a method based on the ideas of classical Proof-of-Work (PoW) for building a distributed registry (Multi-Layer Hash Anchoring, MLHA) based on redundancy of PoW block headers. Each block contains links to the last blocks of all layers, determined by the number of leading hash zeros, which allows you to highlight the sparse skeleton of the waterfall chain. The properties of the waterfall are investigated: uniqueness, determinism, and logarithmic growth of average length. The criterion for the safe deletion of blocks not included in the waterfall is proved, and the possibility of a trustless audit of deleted data is shown. A consensus mechanism based on a lexicographic comparison of waterfall vectors is proposed, consistent with the principle of maximum performance. A single security window has been introduced linking the network hashrate, the minimum mining complexity and the allowable probability of an attack; an upper bound on this probability has been obtained. The simulation results confirm the theoretical estimates. The method does not require a trusted installation, has post-quantum stability and can be used in lightweight decentralized systems.