Convergent Encryption Keys for Multi-Tenant Data Deduplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing encryption methods in multi-tenant data processing systems are incompatible with data deduplication, leading to privacy breaches and inefficient storage due to the inability to deduplicate encrypted data across different users or entities, and face challenges in key management and security.
Innovation Solution
Implement a zero-knowledge architecture using convergent encryption with deduplication perimeters defined by namespaces, generating deterministic encryption keys from the data and namespaces, ensuring that identical data within a perimeter produce the same ciphertext, while different perimeters generate different ciphertexts, thus allowing deduplication without revealing data sharing among tenants.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional encryption is used with user-specific keys, then data privacy is protected, but data deduplication becomes impossible because identical files encrypted with different keys produce different ciphertexts
Solution Approach 1:
The patent introduces a convergent encryption mechanism that acts as an intermediary between traditional encryption and deduplication. Instead of using user-specific random keys, the system derives encryption keys from file content hashes, allowing identical files to produce identical ciphertexts while maintaining encryption. This intermediary approach enables both privacy protection and deduplication capability simultaneously.
Solution Approach 2:
The patent changes the fundamental parameter of encryption key generation from user-specific random values to content-derived deterministic values. By using a hash function on the plaintext file to generate the encryption key, the system ensures that identical inputs always produce identical outputs, enabling deduplication while maintaining security through the one-way nature of hash functions.
2Reliability
If encryption occurs at the client side with user-held keys, then cloud provider cannot access data content, but the provider becomes unable to perform deduplication on uploaded data
Solution Approach 1:
The convergent encryption mechanism enables the encrypted data itself to carry the information needed for deduplication. The encryption key is derived from the file content, so the ciphertext inherently contains the fingerprint information needed for deduplication checks, eliminating the need for separate key management infrastructure and enabling cloud-based deduplication without compromising security.
3Productivity
If convergent encryption is used to enable deduplication, then storage efficiency improves, but controlled information leakage occurs that allows determination of whether plaintexts are identical
Solution Approach 1:
The patent segments the deduplication scope into isolated namespaces, each associated with a specific tenant or user. By applying convergent encryption within segmented namespaces rather than globally, the system prevents cross-tenant information leakage while maintaining deduplication efficiency within each namespace. This segmentation ensures that identical files from different tenants do not reveal information about each other.
Data Source
AI summary
The present invention relates to a method for backing-up encrypted data in a multi-tenant data processing system, comprising the following steps: - Defining a deduplication perimeter per tenant and associates a deduplication namespace DN for each perimeter, - producing a deterministic encryption key KB by applying a hash function f on an object and on the deduplication namespace DN, - encrypting the object using the encryption key KB, - producing a fingerprint of the object by applying a hash function F' on the encrypted object, in order to store only a single instance of identical encrypted objects per deduplication perimeter.