Secured computation: Analyzing sensitive data using homomorphic encryption

Description

Organizations often work with sensitive information such as social security and credit card numbers. Although this data is stored in encrypted form, most analytical operations ranging from data analysis to advanced machine learning algorithms require data decryption for computation. This creates unwanted exposures to theft or unauthorized read by undesirables. Matt Carothers, Jignesh Patel, and Harry Tang explore a use case that prevents fraudulent victimization of customers, implemented in real time, using secured computations enabled through homomorphic encryption. Fraud victimization is one form of economic crime largely executed by deception; the most common form is identification impersonation or using unauthorized credit cards.

Homomorphic encryption gives organizations a secure and easy way to run analytics on data without having to decrypt it by providing computation capability on the encrypted ciphertext, generating an encrypted result which, when decrypted, matches the result of the operations as if they had been performed on the plaintext. Matt, Jignesh, and Harry focus on two use cases. The first case identifies exact identification match and generating alerts if there is mismatch based on predefined attributes. The second use case covers typographical inputs or communication misinterpretation scenarios by identifying proximal identification matches using a combination of Levenshtein distance and Jaro–Winkler distance. And you’ll leave with a background in high-level solution architecture for secure computation, using containers and GPUs to provide real-time information, and the challenges in working with homomorphic encrypted data and performance.