Compression-unimpaired batch-image encryption combining vector quantization and index compression

Abstract

As the network bandwidth has grown rapidly, it has become common to share a large number of still images via the Internet by means of batch transmission. Unfortunately, most existing methods encrypt only a single image, so there is room for improvement in our ability to send batches of images at one time. A complete reconstruction of the methods is sometimes necessary, especially when considering performance criteria. This paper presents a novel batch-image encryption algorithm that combines Vector Quantization (VQ) and additional index-compression process to benefit from their computational efficiency and low transmission bandwidth without affecting the original compression rate. The experimental results show the performance of this new scheme in terms of compression rate and computational cost.

Introduction

The progress of digital technology has brought a great many changes to people’s lives, and digital cameras or cell phones with cameras are regarded virtually as necessities of life. People always have a large number of digital images stored in their memory cards, hard disks or personal discs to be transmitted or shared later over public and possibly hostile networks.

In order to protect the confidentiality of images, the ability to encrypt images is necessary. Although the traditional cryptographic tools, such as DES, AES and RSA [28], have been well defined, they are not suitable for encrypting digital images without modification because the size of digital images is much larger than that of text. Moreover, while a slight distortion in digital images is acceptable for images intended for casual viewing, it is not acceptable with traditional cryptosystems. Hence, it is imperative to design an adaptive encryption algorithm for digital images.

Image encryption methods can be roughly classified into three main groups: full, selective and integrated.

Full encryption was the first approach of traditional encryption algorithms such as AES, DES, and RSA, which encrypted the whole image. However, these methods required a great deal of computation time. Some methods in this category adopt a sophisticated scrambling/chaotic technique to render images into either spatial domain or transformation domain, and some use SCAN patterns [3], [12], [25], [26] and chaos systems [16], [18], [20], [31] to rearrange the pixels of the image or change the pixel values. Encryption using SCAN language involves a sample substitution rule; each SCAN language, defined by a grammar, has a set of basic scan patterns, transformations, and rules with which simple scan patterns are composed to make them complex. Chaos systems usually adopt Torus Automorphism function or Maps, like a permutation function, to disturb the pixels of images. Unfortunately, the computational cost of these methods is also high.

Selective encryption approaches [1], [10], [27] usually transform images into transformation domains such as discrete wavelet transform (DWT) and discrete cosine transform (DCT). In these methods, only some parts of coefficients are scrambled or encrypted. Unfortunately, these encryption algorithms have a potential security drawback [1]: the significant coefficients that are not encrypted are likely to reveal information in the original images. To meet a strict definition of security, the encrypted content must be meaningless and disallow a rough sketch.

This method integrates encryption with compression [6], [9], [21]. While full encryption is secure, its computational cost is high, especially for real-time applications, but selective encryption is usually not secure enough. Multimedia usually distributes in compression because it is straightforward enough to profit from combining encryption and compression.

Researchers have developed a variety of image encryption techniques, and these techniques are progressing in image confidentiality. However, the existing research has addressed only the encryption of a single image and has neglected the fact that people usually distribute digital images in batch transmission.

In addition, because of the limited communication bandwidth, almost all images that travel over the Internet are in a compressed form, like vector quantization (VQ) [17], JPEG, and JPEG2000 [22], to save transmission time and storage space. When we distribute a large number of images, the transmission time and storage space are often an issue, so saving transmission time and storage space becomes a problem. VQ, a powerful lossy image-compression method, is a well known candidate for compression techniques because it is both simple and cost-effective. The encoder can achieve a high compression ratio, and the decoder, who requires only limited information, benefits from fast execution time. VQ works well in some lightweight devices, such as cellular phones. Recently, several studies on VQ [2], [8], [14], [15] and VQ-based applications [5], [6], [7], [9], [19], [29], [30], [32], [33] have been proposed, suggesting its importance.

Combining encryption with compression and image distribution in the form of batch transmission will lead to spectacular opportunities related to batching images. To our best knowledge, the first such scheme was proposed by Chang et al. [6] who encrypted images using VQ and a traditional cryptosystem, but Lee et al. pointed out that the scheme is prone to chosen-plain-image and cipher-image-only attacks [21].

In order to guarantee confidentiality and effective compression, we propose a batch-image encryption system that compresses the VQ indices and uses the modified index-compression method [19]. In this system, index compressing and encryption of a batch of images is done at the same time. The experimental results show the feasibility and advantages of the approach. First, cipher-images are smaller than VQ-compressed images, so the proposed method does not result in any size expansion while adding encryption operations; in fact, it even improves the compression rate. Second, the modified index-compression is adopted to obtain the advantage that the bit rate can be further reduced without loss of image quality. Third, the proposed method has much lower computation costs on the decryption side, a feature that is especially suitable when the receiver is a mobile device equipped with light computational capability.

The rest of paper is organized as follows. Section 2 briefly introduces VQ, the proposed batch-image encryption method is proposed in Section 3, the discussions and experimental results are given in Section 4 Security analysis, 5 Experimental results. Section 6 concludes.