Efficient Homomorphic Encryption for Multikey Compressed Sensing in Lightweight Cloud-Based Image Processing

With the rapid development of cloud storage and privacy computing technologies, users with limited resources are increasingly relying on cloud servers for the storage and computation of their image data. This approach ensures data security and convenient access. However, current image data security-sharing schemes that support privacy computing often face issues such as high bandwidth consumption and significant ciphertext expansion, limiting their applicability. To address these challenges, we propose an innovative multikey compressed sensing lightweight encryption scheme (MCSLE), based on compressed sensing (CS) technology. This scheme is the first to design a multikey conversion algorithm for CS. It allows each sampling end to independently compress the sampled images using different keys. The cloud platform then completes the key conversion and unification. It provides flexible compression sampling capabilities, a robust privacy protection mechanism, and comprehensive support for homomorphic computing. Furthermore, we have designed a specialized image reconstruction algorithm for this scheme. It has undergone in-depth testing in various practical application scenarios, including medical image analysis, fire monitoring, and handwritten text recognition. The experimental results demonstrate that, unlike existing schemes with several tens of times ciphertext expansion, MCSLE can support homomorphic computation at a compression rate of 0.5 while maintaining high-quality image reconstruction.

I. Introduction

With the rapid development of smart devices, the popularity and application of the Internet of Things has become increasingly widespread, which has greatly improved the convenience of People’s Daily lives [1], [2], [3]. In this process, digital images, as a key information medium, play a pivotal role in social media networks, transportation, medicine, and other fields [4], [5], [6]. However, with the rapid expansion of image data, resource-constrained IoT devices, such as sensor nodes, smart home devices, environmental monitoring devices, and smart wearables, are facing enormous data storage and processing pressure. As a result, these devices tend to outsource large amounts of image data to cloud servers for efficient acquisition and in-depth analysis to release pressure on local storage and computing resources [7], [8]. However, open cloud storage environments are susceptible to widespread network attacks, as a large number of images need to be uploaded to cloud servers, and images lacking secure processing face privacy leakage risks. At the same time, due to the large size of image files, the transmission delay and resource consumption will be significantly increased during transmission, which brings challenges to the efficiency and performance of IoT devices. As a result, numerous security-focused solutions have emerged to efficiently process the image data transmitted by IoT devices, for example, schemes based on traditional symmetric cryptography [9], [10], schemes based on public key homomorphic encryption (HE) algorithms [11], [12], and security schemes based on compressed sensing (CS) [8], [13]. Traditional encryption algorithms, such as ECC, RSA, and Paillier, are not well-suited for sensing devices with limited resources. In contrast, security schemes based on CS not only facilitate signal sampling but also achieve compression and encryption, making them widely applicable [14], [15].