Erasing Data from Blockchain Nodes
This addresses a legal and ethical challenge for blockchain operators who need to erase locally stored data without losing full-node participation, offering a novel solution to a domain-specific problem.
The paper tackles the problem of erasing data from blockchain nodes, which is commonly assumed impossible without compromising functionality, by proposing a pragmatic approach called functionality-preserving local erasure (FPLE) that allows full nodes to delete infringing data while continuing to validate most of the blockchain, with a proof-of-concept tool enabling safe compliance with a significantly larger range of erasure requests than existing methods.
It is a common narrative that blockchains are immutable and so it is technically impossible to erase data stored on them. For legal and ethical reasons, however, individuals and organizations might be compelled to erase locally stored data, be it encoded on a blockchain or not. The common assumption for blockchain networks like Bitcoin is that forcing nodes to erase data contained on the blockchain is equal to permanently restricting them from participating in the system in a full-node role. Challenging this belief, in this paper, we propose and demonstrate a pragmatic approach towards functionality-preserving local erasure (FPLE). FPLE enables full nodes to erase infringing or undesirable data while continuing to store and validate most of the blockchain. We describe a general FPLE approach for UTXO-based (i.e., Bitcoin-like) cryptocurrencies and present a lightweight proof-of-concept tool for safely erasing transaction data from the local storage of Bitcoin Core nodes. Erasing nodes continue to operate in tune with the network even when erased transaction outputs become relevant for validating subsequent blocks. Using only our basic proof-of-concept implementation, we are already able to safely comply with a significantly larger range of erasure requests than, to the best of our knowledge, any other full node operator so far.