Privacy-preserving Credit Scoring via Functional Encryption

Abstract: The majority of financial organizations managing confidential data are aware of security threats and leverage widely accepted solutions (e.g., storage encryption, transport-level encryption, intrusion detection systems) to prevent or detect attacks. Yet these hardening measures do little to face even worse threats posed on data-in-use. Solutions such as Homomorphic Encryption (HE) and hardware-assisted Trusted Execution Environment (TEE) are nowadays among the preferred approaches for mitigating this type of threat. However, given the high-performance overhead of HE, financial institutions -- whose processing rate requirements are stringent -- are more oriented towards TEE-based solutions. The X-Margin Inc. company, for example, offers secure financial computations by combining the Intel SGX TEE technology and HE-based Zero-Knowledge Proofs, which shield customers' data-in-use even against malicious insiders, i.e., users having privileged access to the system. Despite such a solution offers strong security guarantees, it is constrained by having to trust Intel and by the SGX hardware extension availability. In this paper, we evaluate a new frontier for X-Margin, i.e., performing privacy-preserving credit risk scoring via an emerging cryptographic scheme: Functional Encryption (FE), which allows a user to only learn a function of the encrypted data. We describe how the X-Margin application can benefit from this innovative approach and -- most importantly -- evaluate its performance impact.