Integrity
Last updated
Last updated
To ensure best action for all platform users is to provide and receive genuine pricing estimates, the protocol is designed to be economically impractical for malicious actors.
In broad strokes, there are two types of malicious actors that could attempt to (A) maximize (undeserved) rewards and (B) skew the output.
Type A Malicious Actors Maliciously high estimates would place Type A Malicious Actors’ estimates further away from the mean estimate unless the malicious submission carries so much RP that substantially outweighs all other submissions combined and becomes the mean. Given RP can only be earned and given the Staking Limits control, attacks by Type A Malicious Actors are highly unlikely. Reward Mechanism allocates rewards among Price Experts based on Z Score of their price estimates and their relative stakes. In particular, Price Experts with Z Score outside of ±0.1 will not be entitled to any reward regardless of their stake. Boosters based on Z Score further allocates increasing share in the reward pools to Price Experts with Z Score closer to 0. Because of these controls, the optimal strategy for Type A Malicious Actors would be to stake just enough RP and maximum LITH with estimates as close to the expected mean as possible.
Type B Malicious Actors Estimates submitted by Price Experts are weighted based on their reputation tokens staked. These reputation tokens (PR) are not transferrable and could only be earned over time through participation on the Lithium ecosystem. New users will be given zero reputation initially and therefore their pricing estimates will carry no weight in until they reach the RP Threshold which entitles them to start staking reputation. In other words, new users are price takers without any influence in pricing enquiries although they are entitled to the same proportionate rewards provided their submissions are within the relevant Z Score reference price estimates. Existing high reputation users, even with substantial RP, are also subject to , requiring them to stake substantial amount of LITH if they wish to stake larger amount of RP to exert a larger market influence. Given the zero reputation for new users, it will be impossible to create new accounts for distorting output price estimate. will also make it financially impractical for organizations to stake substantial RP across multiple accounts (even if they managed to maintain multiple high reputation accounts) given the stake in LITH required to deploy larger amount of RP. In addition to these controls, any attempted manipulation will be further offset or absorbed by the system, the, and participation of other genuine high reputation Price Experts.
Key preventative measures are:
Market Clearing Mechanism The protocol also adopts a two-way pricing estimates model. Ask submissions are required to be higher than the bids submitted by a Price Expert. Because of that, no single Price Expert will be able to produce the indicative market clearing price. Reputations and historical performance (from RP) are also incorporated into the output estimate. Outliers not matched in the will not affect pricing output. Outliers that are cleared with low reputation are also absorbed through the with no or negligible impact on output.
Reputation Must be Earned Although LITH together with precision of estimates determine reward distribution, LITH carries no weight towards the output price estimate. The weight or influence of any input pricing estimate is determined solely by RP or the reputation staked towards a submission.
Encrypted Submission All input pricing estimates will be encrypted prior to submission in order to prevent cheating and late submission advantages. No other Price Expert will have any information about other submissions until the Reveal Stage.
Staking Limits Price Experts will only be entitled to stake RP after achieving and maintaining the RP Threshold. Additional LITH Staking Limit and RP Staking Limit controls are also in place to prevent sybil attacks and any attempts to maintain multiple accounts for distorting outputs.
Positive Reinforcement Once market consensus answer is uncovered, malicious actors' reputation in RP will be slashed, providing positive feedback that will continuously improve the protocol integrity.