Systems and methods for providing a fraud model for proactively preventing attacks on machine learning models

A device, comprising:

• one or more processors configured to:
• receive, from a user, a request for a primary model to generate a prediction;
• cause the request to be processed by the primary model for generation of a primary model prediction;
• process the request, with a fraud model, to generate a fraud model prediction,
• wherein adversarial training is performed on the fraud model based on the primary model and with different training data than used to train the primary model;
• perform a canonical covariance analysis of the primary model prediction and the fraud model prediction to generate results;
• determine metadata and relationships based on the primary model prediction, the fraud model prediction, and user behavior patterns;
• calculate a risk score for the primary model prediction based on the results, the metadata, and the relationships;
• and
• perform one or more actions based on the risk score.

A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising:

• one or more instructions that, when executed by one or more processors of a device, cause the device to:
• perform adversarial training of a fraud model based on a primary model and with different training data than used to train the primary model;
• receive, from a user, a request for the primary model to generate a prediction;
• cause the request to be processed by the primary model for generation of a primary model prediction;
• process the request, with the fraud model, to generate a fraud model prediction;
• perform a canonical covariance analysis of the primary model prediction and the fraud model prediction to generate results;
• determine metadata and relationships based on the primary model prediction, the fraud model prediction, and user behavior patterns;
• calculate a risk score for the primary model prediction based on the results, the metadata, and the relationships,
• wherein the risk score represents a likelihood of anomalies or threats associated with the primary model;
• and
• perform one or more actions based on the risk score.