Криптографічні механізми захисту тестових наборів даних у системах кібербезпеки на основі штучного інтелекту

The article proposes a method for the integrated application of zero-knowledge cryptographic protocols and homomorphic encryption to solve the problem of ensuring the security of test data on which an artificial intelligence cybersecurity system is trained. It justifies the choice of a scheme that is effective in implementation complexity and cryptographic strength for ensuring the confidentiality of the client's test dataset and the training quality of the cybersecurity system provider. An example of the implementation of this scheme for training an email spam filter is provided, which uses a combination of a naive Bayes classifier and a zero-knowledge cryptographic protocol based on a homomorphic cipher.