quevatech/challenge_repoStructured challenge repository offering bounties for breaking a neural-network-based random number generator validated against NIST SP 800-22 and SP 800-90B standards.
🇬🇧 Read in English | 🇹🇷 Türkçe Oku
We have developed a next-generation random number generator that processes Radio Frequency (RF) Noise through a proprietary Neural Network architecture, followed by SHA-3 post-processing. We claim this source provides True Randomness quality, indistinguishable from ideal noise sources.
We are releasing the outputs and the verification reports—but not the source code. The challenge is simple: Prove us wrong.
data/random_sample.bin (~17 MB)Detailed reports can be found in NIST_REPORT.md.
We are offering a bounty to anyone who can successfully "break" this generator.
Reward: Hall of Fame Entry & Certificate of Achievement Task: Given a sequence of $N$ bits from our source, accurately predict the $N+1$ bit with a probability significantly greater than 50% (Statistical significance $p < 0.001$).
Reward: Honorable Mention
Task: Create an algorithm that can distinguish our output from os.urandom or hardware TRNGs with an advantage $> 0.01$.
Reward: Acknowledgment Task: Identify a repeating period or a significant bias in the provided binary sample that was missed by standard NIST tests.
| Date | Researcher | Achievement |
|---|---|---|
| - | - | - |
Powered by Deep Neural Stochastic Processes & Keccak