
Preprint · Submitted to Computers & Security on 23 October 2026
Protecting proprietary website logic
I compared browser-delivered code, encrypted delivery, server execution, extensions, and browser isolation. Code sent to the client remained recoverable in the tested setups; server authorization protected the implementation while leaving permitted use and saved outputs as separate concerns. Learning from outputs worked for a simple function but remained inaccurate for a more complex model. The protected-execution attempt stopped at program loading, so hardware protection was not verified. These experiments establish limits for the tested configurations. Journal submission does not establish peer review or acceptance.
