There is an assumption that cryptographic conditioning reduces the influence of consciousness on RNG. I want to argue that certain types of cryptographic conditioning methods, specifically the method used in CryptoStrong devices, should not reduce mind influence. It actually operates somewhat similarly to the Clairvoyance Psi mode in PsiTrainer.
PsiTrainer includes three Psi modes:
- Clairvoyance
- Mechanism: A set of “target bits” is generated, followed by a set of “psi bits”. The psi bits are then compared to the target bits using AND operator.
- Idea: The user is expected to influence the RNG so that the psi bits match the target bits generated beforehand.
- Precognition
- Mechanism: A set of “psi bits” is generated, followed by a set of “target bits”. The psi bits are then compared to the target bits using AND operator.
- Idea: The user is expected to influence the RNG so that the psi bits match the target bits generated afterwards.
- Psychokinesis
- Mechanism: A set of “psi bits” is generated and compared to a “target bits” of all 1’s using AND operator.
- Idea: The user is expected to influence the RNG to generate as many 1’s as possible.
According to the CryptoStrong whitepaper, the conditioning pipeline works as follows:
- An AES-CTR DRBG (a PRNG) is reseeded 15 times per second using unconditioned bits.
- The final bits are created by XOR’ing fresh unconditioned bits with the AES-CTR DRBG output bits.
When used in a Psychokinesis Psi mode, this construction is somewhat similar to the Clairvoyance Psi mode, where bits influenced by the user are compared to target bits (in this case, the DRBG output bits) that were generated beforehand. The difference is the operation used to compare the bits. PsiTrainer’s Clairvoyance Psi mode uses the AND operator, while CryptoStrong’s conditioning uses XOR.
If the result of “psi bits compared with target bits” is still sensitive to mind influence, then raw/unconditioned bits compared with DRBG bits should still be sensitive.
If the bits’ raw/unconditioned/physical entropy determines their sensitivity to mind influence, using XOR should preserve the original sensitivity, and should be more sensitive than using AND, because XOR preserves entropy while AND reduces entropy to 81.1%:
XOR:
| Bit A | Bit B | A ^ B |
|---|---|---|
| 0 | 0 | 0 |
| 0 | 1 | 1 |
| 1 | 0 | 1 |
| 1 | 1 | 0 |
P(1) = 0.5
P(0) = 0.5
H = -p * log2(p) - (1-p) * log2(1-p)
H = -0.5 * log2(0.5) - 0.5 * log2(0.5)
H = 1.0
AND:
| Bit A | Bit B | A & B |
|---|---|---|
| 0 | 0 | 0 |
| 0 | 1 | 0 |
| 1 | 0 | 0 |
| 1 | 1 | 1 |
P(1) = 0.25
P(0) = 0.75
H = -p * log2(p) - (1-p) * log2(1-p)
H = -0.25 * log2(0.25) - 0.75 * log2(0.75)
H = 0.811
Notes:
-
I’m not arguing that using CryptoStrong-style conditioning to whiten RNG bits can increase an RNG’s sensitivity to mind influence due to apparent randomness quality improvement. I’m only arguing that such conditioning should not reduce the RNG’s sensitivity (if the assumptions in my last two paragraphs are true).
In other words, ComScire CryptoStrong should be as sensitive as ComScire PureQuantum.
-
My argument does not apply to other cryptographic conditioning methods, such as directly outputting DRBG bits seeded by raw/unconditioned bits.