Chaotic Encryption Algorithm With Key Controlled Neural Networks for Intelligent Transportation Systems

Abstract

The security of sensitive information is vital in many aspects of multimedia applications such as Intelligent Transportation Systems (ITSs), where traffic data collection, analysis and manipulations is essential. In ITS, the images captured by roadside units form the basis of many traffic rerouting and management techniques, and hence, we should take all precautions necessary to deter unwanted traffic actions caused by malicious adversaries. Moreover, the collected traffic images might reveal critical private information. Consequently, this paper presents a new image encryption algorithm, denoted as ChaosNet, using chaotic key controlled neural networks for integration with the roadside units of ITSs. The encryption algorithm is based on the Lorenz chaotic system and the novel key controlled finite field neural network. The obtained cryptanalysis show that the proposed encryption scheme has substantial mixing properties, and thus cryptographic strength with up to 5% increase in information entropy compared to other algorithms. Moreover, it offers consistent resistance to common attacks demonstrated by nearly ideal number of changing pixel rate (NPCR), unified averaged changed intensity (UACI), pixel correlation coefficient values, and robustness to cropped attacks. Furthermore, it has less than 0.002% difference in the NPCR and 0.3% in the UACI metrics for different test images.

Convolutional Neural Network Framework for Encrypted Image Classification in Cloud-Based ITS

“…This section presents the results of encrypted image classification, where the objective is to identify the vehicle type in each image. All images are encrypted using AES [22], [29], Twofish [30], ChaosNet [31], Serpent [32]; and electronic codebook (ECB), cipher block chaining (CBC), output feedback (OFB), and the enhanced cipher block chaining (ECBC) modes of operation [33], [34].…”

“…Moreover, the work did not explore the underlying reasons that allow classification algorithms to classify encrypted information. Also, as the data being secured for this framework are images, appropriate encryption algorithms must be used to protect the privacy of the public who are supplying these images [21].…”

“…Moreover, we demonstrate that the performance of such algorithms can be improved for cloud security network applications. In addition, this article utilizes the ChaosNet algorithm [21] to support image encryption for ITS applications. The ChaosNet algorithm was designed as a part of the Scalable-Enhanced Road Side Unit (RSU), based ITS system [23], which was utilized for capturing all images used in this article.…”

Encryption technique based on chaotic neural network space shift and color-theory-induced distortion

“…The Galois field, denoted GF( , contains a finite number of elements, where these elements are integers mod p . We succinctly discuss some of the Galois field GF( results and invite respected readers to see 16 , 17 for detailed discussion.…”

“…The division is defined likewise except that the first element is multiplied by the multiplicative inverse of the second element. We illustrate the multiplication using an example from 17 —3 F and A 5 are two hexadecimal numbers. …”