[Cryptography] Software proposal for verifiable stateless computing
Daniel Hong
shhong at unifiedh.com
Sat Jul 27 06:07:08 EDT 2019
I am developing a verifiable stateless environment builder solution that may be able to execute application code in a trust-less fashion, with little to no performance penalty.
A brief summary of the proposal:
- No concept of "world state": very BitTorrent-like, just dynamic instead of sharing static data
- Uses a pre-defined set of stateless bytecode modules tied together. This represents the call-and-return relationships between modules, that may be used for code integrity verification. They can run in parallel on different devices as long as their parent execution branch is finalized.
- Records all passing parameters and committed state data to the corresponding module & ties them in the form of a chain. This is like a blockchain (Bitcoin, Ethereum etc), but this "state chain" may be flushed as long as all previous parameters and states can be verified in a public way (e.g. using an external blockchain to periodically submit state proofs).
- Run a recursive tree search in case of a parameter and/or state mismatch. Re-execute every leaf module under an affected (malicious) state record.
- Device specific public keys sign corresponding state chain blocks. This is used to encourage honest work & penalize malicious devices.
Unlike a blockchain, state chains are not intended to be publicly verifiable. This approach decreases the computation work required to "finalize" every state transition on a linear blockchain. We can always rely on an external chain when we need it, such as proving the identity of an application or using token contracts.
Currently working on a very simple proof-of-concept to demonstrate the whole idea (source code will be available once it's working).
The paper for this proposal is available at: http://papers.unifiedh.com/Gemer/gemer.pdf. I will be revising and updating the paper as development progresses (this is a very early working draft and not finalized yet).
Any feedback will be greatly appreciated. Thanks.
Daniel Hong
shhong at unifiedh.com
More information about the cryptography
mailing list