Randomness FAQs
On-chain randomness questions and answers
Last updated
Was this helpful?
On-chain randomness questions and answers
Last updated
Was this helpful?
Yes! it is activated on the POA Core and the Sokol Testnet. The RandomAura contract provides on-chain random numbers.
POA Core Contract:
Sokol Contract:
Currently, randomness is created through a -like process, where validators commit the hash of a random number to the chain, then reveal that number later on. The revealed number is XORd with a previous random seed, creating a new random seed. For more details, see
In the future, we will have
No! Random values are only created during the reveals phase, which occurs every 20 blocks and continues for a period of 20 blocks (note this value is configurable).
A complete collection phase on both POA Core & Sokol is currently set to 40 blocks. The first half (the first 20 blocks) is called the commit phase, where random number hashes are committed by validators. The second half (the second 20 blocks) is the reveal phase, where numbers are revealed and added to the currentSeed getter.
Entropy increases throughout the reveal phase, and the final number revealed is the most secure. Applications requiring secure randomness should retrieve the currentSeed from the final block of a reveal phase or during a commit phase.
The value is contained in the currentSeed getter. It is important to check the phase (commit or reveal) and determine when the value is created. Details are available on the page.
Business logic actions that require randomness should not be allowed during the reveal phase. In addition, randomness can never be guaranteed for block N + 1, only for some block between N + commitRoundLength and N + 2*commitRoundLength.
It is possible to create an on-chain PRNG where the currentSeed value is used to seed a generator. However, as soon as the seed is known, the whole sequence is known! To add additional entropy, the seed may be salted with the block hash, however this method is still not considered secure.
While secure, there are considerations to keep in mind. They have to do with malicious validators who may choose to manipulate the outcome by not revealing their number. Validators cannot change the number they have committed, but they can choose to not commit or not reveal a number.
This means that during the reveal phase, a validator can effectively choose between 2 numbers, either the current number or the new one that will be created when they reveal their number. If an application uses the final number of the reveals phase, only the final validator can make this choice, limiting the scope of this issue.
To discourage skipping, validators who skip too often (or skip at the end of an epoch) will be reported as malicious. In POSDAO (a proof-of-stake algorithm that may be implemented in the future), malicious validators will be banned from the protocol for 90 days and their STAKE frozen. For now, we will monitor the network for any malicious behavior and validators can determine and vote on consequences.
Since validators on Sokol and POA are known individuals staking their reputation, this is not as much a concern as it would be in a permissionless network.
YES! When we move to HoneyBadger BFT, reliable random numbers will be produced per block via threshold signatures.
Using this approach, validators will signal their approval of a block by providing a portion of a signature (a signature share) rather than the entire signature. Once a predetermined number of shares are received by the algorithm (the threshold), they are combined to create a single signature which cannot be known beforehand. Because this number is secret until it is revealed, it can be used as a random number. A special property of this algorithm is that any combination of validators can collaborate to create the same final signature.
A simple way to turn a single seed into multiple numbers is to use hash(currentseed+0) , hash(currentseed+1), hash(currentseed+2), etc., or something similar (e.g. hash(currentseed+previousHash)). Limitations of this method (regarding security and speed) are discussed here:
For more on HoneyBadger BFT, see