Homomorphic Encryption AI Model Conversion via Knowledge Distillation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI models transmit user data and responses in plaintext, making them vulnerable to leakage by server administrators or third parties, and converting these models to homomorphic encryption AI models is computationally challenging.
Innovation Solution
A server device and method that convert a plaintext AI model into a homomorphic encryption AI model by receiving a reference AI model, performing a knowledge distillation task to acquire a lightweight AI model friendly to homomorphic encryption, and encrypting the data used by the plaintext AI model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a plaintext AI model is converted to a homomorphic encryption AI model, then security is improved, but computational complexity increases
Solution Approach 1:
The conversion process is divided into two distinct stages: first performing knowledge distillation to create a lightweight AI model, then converting that lightweight model to homomorphic encryption. This segmentation reduces the overall computational complexity compared to directly converting the original large model.
Solution Approach 2:
Knowledge distillation is performed as a preliminary action before the homomorphic encryption conversion. By first creating a simplified lightweight model that captures the essential functionality of the original model, the subsequent encryption process becomes computationally more manageable.
2Measurement precision
If a large reference AI model is used, then model accuracy is improved, but conversion time increases
Solution Approach 1:
Instead of directly encrypting the large reference AI model, the system creates a copied simplified version through knowledge distillation. This lightweight copy retains the essential predictive capabilities while being much more efficient for homomorphic encryption conversion and execution.
Solution Approach 2:
The knowledge distillation process changes the parameters of the AI model by creating a lightweight version with fewer layers and parameters. This parameter transformation maintains adequate accuracy for the intended application while dramatically reducing the computational burden of homomorphic encryption conversion.
Data Source
AI summary
Disclosed is a server device. The device includes: a communicator; a memory; and a processor, wherein the processor is configured to receive a reference artificial intelligence (AI) model of an external device through the communicator and store the received reference AI model in the memory, acquire a plaintext AI model friendly to homomorphic encryption by performing a knowledge distillation task based on the reference AI model for a lightweight AI model designed to operate homomorphic encryption efficiently compared to the reference AI model, and convert the plaintext AI model into a homomorphic encryption AI model by encrypting data used by the plaintext AI model. Accordingly, the device may easily provide the homomorphic encryption AI model.