Novusphere: A Censorship Resistant Liability-free Database
Abstract. A purely peer-to-peer database allows records to be stored in its dataset with state transformations preserved via decentralized consensus ensuring the records were resistant to censorship. As no individual party has the authority to remove a record from the agreed upon dataset, no party can be held completely liable for any record in the dataset. The decentralization of the administrative authority to delete records and the recording of all state transitions ensures the existence of a significant barrier to censorship. Since moderation to keep data out is significantly difficult, the only form of moderation left is voluntary opt-in.
Censorship on the internet is a problem where large social media giants such as Facebook, Twitter, Reddit, etc. control the power to heavily influence the user base including political events such as the election of a president. Through censorship and thought policing, narratives can easily be crafted and agendas easily pushed. Existing social media platforms and content sharing mediums work sufficiently well for many people and use cases. However, they suffer from the inherent weakness that the administrators of such platforms can exploit their administrative rights to police or be forced to police content.
What is needed is a base system where information cannot be tampered with or censored. With such a system, any party can freely share content including thoughts, ideas, and media, or arbitrary files. In this paper, we propose a solution to mitigate censorship via administrative abuse in traditional server-side database model using a peer-to-peer blockchain-based network for content-agnostic communication. The content is secure as long as censorship resistance of the underlying blockchain is secure. We propose the use of a delegated proof of stake (DPoS) model,,,, secured via honest users controlling honest block producers via greater committed stake than any colluding malicious groups. In the event the system is ever fully compromised, the honest users have the option to fork the system’s state to a new chain void of malicious actors’ stake.
The monetary unit or token of the platform (atmos) must first be vested to have a stake in the system. Vested atmos referred to as v-atmos are a direct representation of your stake in the system. A user can have multiple vestings occuring at the same time.
When a user requests their atmos be vested, the amount becomes unspendable and will transfer over time from the atmos locked in the vested transaction to v-atmos until all atmos has become v-atmos. Likewise, a user can request their v-atmos be divested, which will initially deducted the amount from their total v-atmos. Over time the v-atmos will be released back to the user as atmos, effectively the reverse of vesting.
The system is then governed by stakeholders voting on various things such as DAO funding proposals and development decisions.
In the early stages of Novusphere and atmos on EOS, there will be no inflation and supply will be a fixed number. Inflation will later be introduced after the public database system has been fully implemented.
Blockchain Architecture And Requirements
While blockchains modeled after Bitcoin and Ethereum do not meet our requirements, Steem’s white paper proposed “Bandwidth Instead of Micropayment Channels” as a solution to allowing zero fee transactions. A similar model is proposed in EOS under “Token Model and Resource Usage” which allocates resources based on an individual's EOS holdings.
To further incentivize content being added to the database, we propose a system in which users (requesters) can create bounties rewarding atmos for specific requests which can then be fulfilled by other users (bounty hunters) to claim the reward. To mediate this system, stakeholders elect a panel of judges to help settle any disputes that may arise.
For example, two bounty hunters A and B submitted content attempting to claim the bounty. The requester has chosen B to receive the reward, however A feels cheated since they were first to submit a satisfactory request. A opens a case with the judges by offering collateral equal to the “intervention fee” which is a percentage of the total bounty reward. Should there be multiple cases opened regarding the same bounty, each person must offer their own collateral equal to that of the intervention fee upon opening the case, but once settled the intervention fee will be deducted evenly from each party with an open case meaning each person’s entire collateral is not consumed.
Judges must opt-in first and accept a case and must also vote with the greatest majority to be paid subsidy from the intervention fee. As judges must vote in a blind mannerism, most people will vote what they believe to be correct or the “honest” outcome expecting the other judges to do the same. There is no financial motive to be dishonest.
Varying degrees of punishments to judges can occur if:
The database state is derived by the application layer from the blockchain layer. It is important to make the distinction that the application layer is agnostic to the underlying blockchain layer and only cares about data retrieval, storage and access. The blockchain layer must make reference to IPFS hashes of which the application layer monitors and reconstructs into a database of state transitions. The application layer downloads the json object referenced by the IPFS hash and then indexes it to a database model. Updates to an object can be made by providing a state transition json object which includes the changes in the object. Branches of an object can be made if someone who is not the current owner of an object creates a state transition.
When an object is retrieved from the database by, it will also return a history of state transitions that have occured by the same publisher of the object trying to be retrieved.
For example, Bob has an object which he has committed the initial state and 3 transitions to. Alice creates a branch at B2, and adds 2 transitions.
B1 ← B2 ← B3 ← B4
← A3 ← A4
When the object B1, B2, B3 or B4 is attempted to be retrieved either the most recent state “B” with all transitions applied can be retrieved or all individual transitions (B1, B2, B3, B4).
Likewise, when the object A3 or A4 is retrieved either the most recent state “A” with all transitions applied can be retrieved or all individual transitions (B1, B2, A3, A4).
This provides an immutable history of an object while still allowing a mutable object to exist.
Efficient Database Reconstruction
Reconstructing the state of a database from point-zero may be trivial early on, but as the state database grows in numbers of rows this may be a slow process. To circumvent the need to download all data from the blockchain, a light-weight model can be implemented which uses snapshots every 1,000,000 rows. Snapshots can be downloadable via public database maintainers and used as a way to quickly download and sync the database state. A snapshot’s hash should be published on chain, and public database maintainers should vote in agreeance with the hash. One can verify a snapshots validity by comparing it to the hash on chain, as well as examining the hash provided by different public database maintainers.
The combination of a decentralized blockchain offering censorship resistance, content indexing for consumption, bounty system for creating desired content, and judge system for reviewing content start to approach a sustainable mutually beneficial system. In addition it creates opportunities and incentives for users to become part of the system be it as a block producer, voter, requester, bounty hunter, or a judge.
Figure. Graphical schematic of the mechanism for content growth of the system.
 "DPOS Consensus Algorithm - The Missing White Paper — Steemit." Accessed May 4, 2018. https://steemit.com/dpos/@dantheman/dpos-consensus-algorithm-this-missing-white-paper.
 "In Defense of Consortium Blockchains — Steemit." Accessed May 4, 2018. https://steemit.com/eos/@dan/in-defense-of-consortium-blockchains.
 "Proof of Stake versus Proof of Work - BitFury." Accessed May 4, 2018. http://bitfury.com/content/5-white-papers-research/pos-vs-pow-1.0.2.pdf.
 "Response to Cosmos white paper's claims on DPOS security — Steemit." Accessed May 4, 2018. https://steemit.com/steem/@dantheman/response-to-cosmos-white-paper-s-claims-on-dpos-security.
 “EOS.IO Technical White Paper v2” Accessed May 7, 2018.