IIB–CPE: Inter and Intra Block Processing-Based Compressible Perceptual Encryption Method for Privacy-Preserving Deep Learning

Abstract

Perceptual encryption (PE) of images protects visual information while retaining the intrinsic properties necessary to enable computation in the encryption domain. Block–based PE produces JPEG-compliant images with almost the same compression savings as that of the plain images. The methods represent an input color image as a pseudo grayscale image to benefit from a smaller block size. However, such representation degrades image quality and compression savings, and removes color information, which limits their applications. To solve these limitations, we proposed inter and intra block processing for compressible PE methods (IIB–CPE). The method represents an input as a color image and performs block-level inter processing and sub-block-level intra processing on it. The intra block processing results in an inside–out geometric transformation that disrupts the symmetry of an entire block thus achieves visual encryption of local details while preserving the global contents of an image. The intra block-level processing allows the use of a smaller block size, which improves encryption efficiency without compromising compression performance. Our analyses showed that IIB–CPE offers 15% bitrate savings with better image quality than the existing PE methods. In addition, we extended the scope of applications of the proposed IIB–CPE to the privacy-preserving deep learning (PPDL) domain.

Comprehensive Analysis of Compressible Perceptual Encryption Methods—Compression and Encryption Perspectives

“…However, these schemes have a prerequisite of a color image as an input, for example, to achieve a large number of blocks, the individual color component processing (Extended CPE methods) and the pseudo-grayscale image representation (PGS-CPE methods) are only possible when the input is a color image. This advantage of these methods diminishes when the input image is a grayscale image with only one channel [49]. To overcome this limitation, Ahmad et al proposed in [34,49,50] an inside-out transformation function that performs the rotation and inversion step on a sub-block level.…”

“…This advantage of these methods diminishes when the input image is a grayscale image with only one channel [49]. To overcome this limitation, Ahmad et al proposed in [34,49,50] an inside-out transformation function that performs the rotation and inversion step on a sub-block level. Compared to the CPE methods that transform an entire block, these methods have a larger keyspace size for grayscale image processing.…”

“…An IIB-CPE scheme is proposed in [34,49,50] to expand the keyspace of Color CPE methods. The core idea is to perform sub-block processing.…”

“…In similar work presented in [63], the authors used an Extended CPE method for face recognition in a color image dataset. Besides face recognition tasks, CPE-based privacy-preserving image classification has been performed in [49,[64][65][66]. Specifically, in [64], the authors implemented an isotropic network such as vision transformers with the Color CPE scheme for natural image classification.…”

“…In [65], the authors implemented four different extensions of IIB-CPE and analyzed their effect on a CNN model's accuracy. The same authors implemented a CNN-based model with a IIB-CPE scheme for natural image classification in [49] and for COVID-19 diagnosis in chest X-ray images in [66].…”