Developer Tests Frontier LLM's Ability to Extract Secrets from Obfuscated App Builds

A developer ran an experiment to assess how difficult it would be for an outsider to recover secrets from a hardened application build, using a frontier large language model as the investigator. The app's configuration values are XOR-encoded, with working keys derived at runtime from a 32-byte master key that ships inside the binary but is never stored as a readable string. The developer acknowledged upfront that this constitutes obfuscation, not cryptographic protection, since the binary contains all material needed for decoding. Key hardening measures included removing readable strings and decorrelating encoded values to prevent clustering by shared prefixes. The experiment focused not just on whether the LLM could recover secrets, but on how to critically evaluate its findings and gauge confidence in a failed recovery attempt.