Cryptographic token logit biasing, statistical z-score detection & evasion attack testbench
The written detector compares green count to token count times the assigned green fraction, then divides by square root of count times fraction times one minus fraction. With an illustrative one hundred evaluated tokens and fraction one quarter, expected count is twenty-five and standard deviation about four point three three. Thirty-five green tokens give z about two point three zero nine; forty-five give four point six one nine. Both bars depict z on one hundred thirty pixels per unit. These illustrative counts explain the algebra, not a measured false-positive rate or proof of authorship. The context window leaves that many initial tokens unscored. For an illustrative twenty-token passage, window one gives nineteen evaluated tokens and window three gives seventeen. Both bars show evaluated counts on one thirty-pixel scale. Each scored token uses the preceding window of token strings plus its own token and key for the source hash. The displayed protocol and compliance labels do not make this simple hash a validated cryptographic implementation or legal certification. Detection in the saved lab compares its computed z against a source threshold rather than verifying who wrote the passage. As an illustrative count calculation, one hundred evaluated tokens at fraction one half have expectation fifty and standard deviation five. Sixty green tokens yield z two; seventy yield z four. Both bars use one hundred fifty pixels per z unit. Editing words or a key can change the hash count. An attack uses authored substitutions and sampling, so its displayed retained score remains a simulation outcome rather than a general robustness guarantee.