Adaptively secure streaming functional encryption system for privacy-preserving computation on evolving datasets

Background

The field of cryptography increasingly focuses on privacy-preserving computations for continuous data streams, such as live sensor feeds and evolving medical records. In these environments, data owners must share specific insights without exposing raw information. Functional encryption addresses this by allowing a keyholder to learn only the output of a designated function evaluated on encrypted data. As digital infrastructures rely more on real-time analytics, there is a critical need for cryptographic frameworks capable of providing fine-grained access to continuous data streams as they are generated, ensuring sensitive information remains protected while actively utilized.

Despite this need, current approaches face significant limitations in dynamic environments. Traditional functional encryption requires the entire dataset to be fully assembled at the time of encryption, making it inherently incompatible with continuous data streams. While some streaming-oriented models exist, they suffer from critical vulnerabilities, achieving only limited, semi-adaptive, or function-selective security. Crucially, these prior methods fail to protect data when adversaries interleave requests for function keys and ciphertexts in an unpredictable order. Consequently, existing frameworks are insufficient for real-world applications that demand robust security guarantees against sophisticated, dynamic threats.

Technology

This streaming functional encryption (sFE) system enables secure, incremental, and function-specific computations on continuous datasets like live sensor feeds. The architecture integrates a modular suite of routines for setup, streaming encryption, key generation, and decryption. It leverages sophisticated cryptographic primitives, including indistinguishability obfuscation to hide decryption logic, splittable signatures for authentication, and cryptographic iterators to manage secure state transitions. Additionally, a unique bootstrapping mechanism transitions the framework from a single-key, secret-key model to a robust, multi-key, public-key environment capable of processing large-scale data streams in real time.

The technology is highly differentiated by its achievement of full adaptive security for general polynomial-sized circuits, overcoming major limitations in traditional cryptographic models. Unlike standard functional encryption that requires an entire dataset to be present before encryption, this solution efficiently processes data dynamically as it is generated. While previous streaming encryption schemes only offered limited, semi-adaptive security, this framework remains completely secure even if adversaries interleave requests for function keys and ciphertexts in an arbitrary order. This robust protection is accomplished through a novel non-black-box “gluing” technique that combines message-selective and function-selective schemes. Consequently, it provides unprecedented privacy-preserving capabilities for dynamic machine learning.

Benefits

• Enables secure, privacy-preserving computations on continuous, real-time data streams (such as live sensor feeds or medical records) without exposing the underlying raw data.
• Provides full adaptive security, ensuring robust protection even when attackers request function keys and ciphertexts in unpredictable, interleaved orders.
• Supports dynamic key management, allowing new function-specific decryption keys to be issued seamlessly in the middle of an active data stream without disrupting operations.
• Delivers high throughput and low latency, making the system highly scalable for processing large-volume, high-speed data environments.
• Enforces fine-grained access control by ensuring keyholders can only extract specific, authorized computational outputs from the encrypted data.
• Facilitates advanced, real-time applications such as privacy-preserving machine learning on evolving datasets, confidential cloud processing, and secure live video analysis.

Applications

• Privacy-preserving machine learning
• Secure live video stream analysis
• Confidential cloud data processing
• Secure medical records research
• Secure live sensor data analysis

Additional info

This technology provides adaptively secure streaming functional encryption for incremental computation on evolving datasets. It integrates indistinguishability obfuscation, splittable signatures, and cryptographic iterators to hide decryption logic and manage state transitions. By merging message-selective and function-selective schemes, the architecture ensures security regardless of request order. A bootstrapping mechanism employing function-private encryption enables transition to public-key environments, facilitating secure, real-time analysis of large-scale, privacy-sensitive data streams.