Skip to main content

Security Explained

Entropy Mixing​

To ensure security of the final random value generated, your seed is always mixed with additional on-chain entropy before being used as input to the VDF.

seed = H(requestId, user_seed, prevrandao, prev blockhash)

When combined with a properly designed Profile, this ensures that no participant can predict the random result early enough to manipulate the outcome of your application.

Delay Parameter​

DeRand mixes the user-provided seed with entropy from previous blocks.

In addition, the VDF is tightly coupled to a time-based delay. We therefore define a conservative upper bound on the time elapsed between the current block and the previous block:

delta_time = current_block_timestamp - checkpoint_block_timestamp

Here, checkpoint_block_timestamp refers to the timestamp of the latest block observed by the requester at the time of submitting the request (not the previous block of the request itself).

This delta_time represents the maximum time window available for any adversary to attempt to influence or attack the request.

DeRand uses this time window to derive the VDF iteration count:

delay = delayFactor * ceil(delta_time / profile.delayTime)
if delay > maxDelay:
  revert

T = delay * profile.delayScale

Where:

  • delayFactor increases the safety margin against participants with faster hardware or against validator collusion.
  • maxDelay protects the requester from unexpectedly large delays and fees caused by unusually long block intervals.
  • T is the final VDF iteration count.

Profile Selection​

VDFs require an unknown-order group, which prevents adversaries from exploiting algebraic shortcuts to evaluate the computation significantly faster than the intended sequential process.

Class groups are a widely studied construction for unknown-order groups and have the advantage of not requiring a trusted setup. Profiles based on class groups with larger discriminants (D) generally provide stronger resistance against known attacks on the underlying group structure.

When selecting a Profile:

  • Prefer Profiles that use class groups or other well-studied unknown-order group constructions.
  • Profiles with larger discriminants generally provide stronger security.
  • Choose a Profile with fees that are appropriate for the desired level of prover participation and service reliability.

The discriminant size, security assumptions, fees, and other parameters are documented in each Profile's details.