Secure Cox Regression Testing for Privacy-Preserving Analysis

The patent segments the statistical test process into multiple components that can be executed in encrypted form. Instead of performing tests on decrypted data, the system divides the test calculations into separate operations (score vector computation, Hessian matrix computation, Wald test statistics) that can be performed on encrypted inputs, thereby preserving both privacy and test capability.

The patent introduces encrypted intermediate representations (encrypted score vectors, encrypted Hessian matrices, encrypted test statistics) as mediators between the private data and the test results. These intermediaries allow statistical tests to be performed without exposing the underlying sensitive data, resolving the contradiction between privacy preservation and test execution.

The patent changes the computational parameters from operating on plaintext data to operating on ciphertext data. By transforming the entire computation pipeline to work with encrypted representations, the system maintains data confidentiality while enabling complex statistical operations through cryptographic transformations rather than traditional numerical methods.

The patent substitutes traditional mechanical computation (direct calculation on plaintext) with cryptographic computation (calculation on ciphertext). This replacement allows statistical operations to be performed through encrypted arithmetic and cryptographic primitives, eliminating the need to decrypt data while still achieving the computational goals.