Functional Encryption for Vertical Federated Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for vertical federated learning face inefficiencies due to reliance on computationally expensive additive homomorphic encryption and garbled-circuit techniques, are model-specific, limited to two participants, require multiple iterations of communication, and are vulnerable to inference attacks, making them slow, inefficient, and insecure.
Innovation Solution
Implementing a two-phase non-interactive secure aggregation approach using hybrid functional encryption techniques, which generates and distributes single-input and multi-input functional encryption keys to enable secure and efficient aggregation of model updates and datasets without inter-participant communication, reducing vulnerability to inference attacks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If additive homomorphic encryption and garbled-circuit techniques are used for vertical federated learning, then privacy protection is improved, but computation efficiency deteriorates
Solution Approach 1:
The patent changes the cryptographic parameters and techniques from traditional homomorphic encryption and garbled circuits to functional encryption with inner product computation. This parameter change enables efficient computation while maintaining privacy guarantees through the mathematical properties of functional encryption, resolving the contradiction between privacy protection and computation efficiency.
Solution Approach 2:
The patent substitutes the mechanical cryptographic operations of homomorphic encryption and garbled circuits with a more efficient functional encryption mechanism that uses inner product computation. This substitution replaces complex cryptographic mechanics with a streamlined approach that achieves the same privacy goals with significantly lower computational overhead.
2Reliability
If garbled-circuit-based secure multi-party computation is implemented, then privacy protection is improved, but communication efficiency deteriorates
Solution Approach 1:
The patent changes the communication protocol parameters from multi-iteration garbled-circuit exchange to single-iteration functional encryption key distribution. This parameter change reduces communication rounds from multiple back-and-forth iterations to a single efficient key distribution phase, dramatically reducing communication overhead while preserving privacy.
3Adaptability or versatility
If existing vertical federated learning systems are extended to more than two participants, then adaptability is improved, but system complexity and communication overhead increase excessively
Solution Approach 1:
The patent implements a universal functional encryption framework that handles any number of participants through a standardized key distribution mechanism. The system uses a single public key for all participants and generates individual secret keys through a unified process, making the system scalable and adaptable to any participant count without proportionally increasing complexity.
Solution Approach 2:
The patent segments the participant management into independent key pairs (public key for all, individual secret keys for each participant). This segmentation allows each participant to be added or removed independently through simple key distribution without reconfiguring the entire system, reducing complexity while improving adaptability to dynamic participant groups.
4Ease of operation
If participants communicate directly with each other in vertical federated learning, then model training is facilitated, but security against inference attacks deteriorates
Solution Approach 1:
The patent introduces functional encryption as an intermediary layer between participants. Instead of direct communication that exposes data to inference attacks, the system uses encrypted functional keys as mediators that enable model training computations while preventing direct observation of sensitive data by other participants, thus eliminating inference attack vulnerabilities.
5Ease of manufacture
If model-specific approaches are used for vertical federated learning, then implementation simplicity is improved, but versatility deteriorates
Solution Approach 1:
The patent implements a universal functional encryption framework that works across different machine learning models through standardized inner product computation. The same cryptographic mechanism supports various model types by operating on their respective parameter vectors, providing both implementation simplicity through a unified approach and versatility across different model architectures.
Data Source
AI summary
Systems and techniques that facilitate universal and efficient privacy-preserving vertical federated learning are provided. In various embodiments, a key distribution component can distribute respective feature-dimension public keys and respective sample-dimension public keys to respective participants in a vertical federated learning framework governed by a coordinator, wherein the respective participants can send to the coordinator respective local model updates encrypted by the respective feature-dimension public keys and respective local datasets encrypted by the respective sample-dimension public keys. In various embodiments, an inference prevention component can verify a participant-related weight vector generated by the coordinator, based on which the key distribution component can distribute to the coordinator a functional feature-dimension secret key that can aggregate the encrypted respective local model updates into a sample-related weight vector. In various embodiments, the inference prevention component can verify the sample-related weight vector, based on which the key distribution component can distribute to the coordinator a functional sample-dimension secret key that can aggregate the encrypted respective local datasets into an update value for a global model.