International business machines corporation (20240243898). VERIFYING REMOTE EXECUTION OF MACHINE LEARNING INFERENCE UNDER HOMOMORPHIC ENCRYPTION USING PERMUTATIONS simplified abstract
VERIFYING REMOTE EXECUTION OF MACHINE LEARNING INFERENCE UNDER HOMOMORPHIC ENCRYPTION USING PERMUTATIONS
Organization Name
international business machines corporation
Inventor(s)
Eyal Kushnir of Kfar Vradim (IL)
Ramy Masalha of Kafr Qari (IL)
Nir Drucker of Zichron Yaakov (IL)
VERIFYING REMOTE EXECUTION OF MACHINE LEARNING INFERENCE UNDER HOMOMORPHIC ENCRYPTION USING PERMUTATIONS - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240243898 titled 'VERIFYING REMOTE EXECUTION OF MACHINE LEARNING INFERENCE UNDER HOMOMORPHIC ENCRYPTION USING PERMUTATIONS
Simplified Explanation: The patent application describes a technique to remotely identify potential compromise of a service provider that performs homomorphic inferencing on a model using real data samples.
Key Features and Innovation:
- Generation of first and second permutations of trigger samples for real data samples.
- Homomorphic inferencing on the model at least twice in a secret permutated way.
- Use of a tile tensor data structure to package real data samples prior to encryption.
- Determination of service provider compromise based on results from inferencing.
- Mitigation action taken upon detection of compromise.
Potential Applications: This technology can be applied in cybersecurity, data privacy, and secure machine learning applications.
Problems Solved: The technology addresses the challenge of identifying potential compromise in a service provider performing homomorphic inferencing on sensitive data.
Benefits: Enhanced security and privacy in homomorphic inferencing processes, early detection of service provider compromise, and improved data protection.
Commercial Applications: Potential commercial applications include secure cloud computing services, data analytics companies, and organizations handling sensitive data.
Prior Art: Prior research in homomorphic encryption, secure inferencing techniques, and data privacy methods may be relevant to this technology.
Frequently Updated Research: Stay updated on advancements in homomorphic encryption, secure machine learning, and data privacy regulations.
Questions about the Technology: 1. How does the use of tile tensor data structure enhance the security of real data samples during homomorphic inferencing? 2. What are the potential implications of early detection of service provider compromise in secure machine learning applications?
Original Abstract Submitted
a technique to remotely identify potential compromise of a service provider that performs homomorphic inferencing on a model. for a set of real data samples on which the inferencing is to take place, at least first and second permutations of a set of trigger samples are generated. every set of samples (both trigger and real samples) are then sent for homomorphic inferencing on the model at least twice, and in a secret permutated way. to improve performance, a permutation is packaged with the real data samples prior to encryption using a general purpose data structure, a tile tensor, that allows users to store multi-dimensional arrays (tensors) of arbitrary shapes and sizes. in response to receiving one or more results from the he-based model inferencing, a determination is made whether the service provider is compromised. upon a determination that the service provider is compromised, a given mitigation action is taken.
