Showing papers by "Sudhish N. George published in 2014"
20 Feb 2014
TL;DR: From experimental analysis, it is proven that the proposed system provides better performance than its counterparts and can provide security maintaining the robustness to noise of the CS system.
Abstract: In this paper, a novel approach for generating the secure measurement matrix for compressive sensing (CS) based on linear feedback shift register (LFSR) is presented. The basic idea is to select the different states of LFSR as the random entries of the measurement matrix and normalize these values to get independent and identically distributed (i.i.d.) random variables with zero mean and variance \(\frac{1}{N}\), where N is the number of input samples. The initial seed for the LFSR system act as the key to the user to provide security. Since the measurement matrix is generated from the LFSR system, and memory overload to store the measurement matrix is avoided in the proposed system. Moreover, the proposed system can provide security maintaining the robustness to noise of the CS system. The proposed system is validated through different block-based CS techniques of images. To enhance security, the different blocks of images are measured with different measurement matrices so that the proposed encryption system can withstand known plaintext attack. A modulo division circuit is used to reseed the LFSR system to generate multiple random measurement matrices, whereby after each fundamental period of LFSR, the feedback polynomial of the modulo circuit is modified in terms of a chaotic value. The proposed secure robust CS paradigm for images is subjected to several forms of attacks and is proven to be resistant against the same. From experimental analysis, it is proven that the proposed system provides better performance than its counterparts.
49 citations
TL;DR: It is experimentally proved that the proposed encryption system can maintain the robustness to noise of the compressive sensing system and is subjected to several forms of attacks and is proved to be resistant against all.
Abstract: In this paper, a novel approach for secure compressive sensing of images based on multiple one dimensional chaotic maps is proposed. The basic idea is to perform the random selection of a combination of two one dimensional chaotic maps to generate a random stream. One or more values from the random stream are used to generate each normal value in the random measurement matrix for compressive sensing. In the proposed approach, eight different one dimensional chaotic maps are used. For one measurement matrix generation, two of them are randomly selected based on a secret key. The number of iterations and the initial states of the chaotic maps are also decided by external secret keys. The chaotic output of iterations of the two selected chaotic maps are XORed to generate a new chaotic value so that the measurement matrix so generated can withstand known plaintext attack. The block-based compressive sensing (BCS) of images is adopted to validate the proposed system. Further enhancement in the security of the proposed system for BCS of images can be obtained by using different measurement matrices for different blocks of images. An algorithm for generating multiple measurement matrices is also presented in this work. It is experimentally proved that the proposed encryption system can maintain the robustness to noise of the compressive sensing system. The proposed robust encryption system is subjected to several forms of attacks and is proved to be resistant against all.
21 citations
17 Feb 2014
TL;DR: The proposed encryption approach can be applied to any audio coding system, which maintains the standard compatibility and is proved that it can provide better security than the existing ones with very less system complexity.
Abstract: In this paper, a novel method of encrypting the encoded audio data based on LFSR based key stream generators is presented. The alternating step generator (ASG) is selected as keystream generator used for this application. Since the ASG is vulnerable to improved linear consistency attack, it is proposed to incorporate some nonlinearity with the stop/go LFSRs of the ASG so that the modified ASG can withstand the same. In the proposed approach, the selected bits of each frame of the encoded audio data is encrypted with the keystream generated by the modified ASG. In order to overcome known plaintext attack, it is proposed to use different keystreams for different frames of the audio data. The long keystream generated from the modified ASG is divided into smaller keystreams so that it can be used as the different keystreams for different frames of audio data. The proposed encryption approach can be applied to any audio coding system, which maintains the standard compatibility. The number of encrypted bits control the degree of degradation of the audio quality. The performance of proposed encryption method is verified with MP3 coded audio data and is proved that it can provide better security than the existing ones with very less system complexity.
9 citations
17 Feb 2014
TL;DR: A novel approach for scrambling the compressively sensed audio data, in which the initial state matrix of CA is constructed by the random bit sequence generated by a linear feedback shift register (LFSR).
Abstract: This paper proposes a novel approach for scrambling the compressively sensed audio data. Two dimensional cellular automata (CA) based scrambling algorithm is proposed, in which the initial state matrix of CA is constructed by the random bit sequence generated by a linear feedback shift register (LFSR). Compressive sensing (CS) offers excellent robustness in the presence of noise and greater compression whereas CA offers excellent scrambling degree. This scrambling scheme ensures efficient channel usage, resistivity to noise, best signal to noise ratio and good scrambling of data. Experimental results confirm the effectiveness of the proposed scheme.
7 citations
13 Mar 2014
TL;DR: The combination of compressive sensing and arnold scrambling provides very high security and ensures efficient channel usage, resistivity to noise, best signal to noise ratio and good scrambling of data.
Abstract: In this paper, a novel idea for scrambling the compressive sensed audio data using two dimensional Arnold transform is presented. In the proposed method, Arnold matrix is constructed by the numbers generated by using a secret key and a logistic map. A key based measurement matrix is used for compressive sensing to avoid the transmission and storage requirement of the matrix and to improve the security. The combination of compressive sensing and arnold scrambling provides very high security and ensures efficient channel usage, resistivity to noise, best signal to noise ratio and good scrambling of data. Experimental results confirm the effectiveness of the proposed scheme.
5 citations
10 Jul 2014
TL;DR: From experiments, it is p roved that the proposed system gives better compression performance and security with minimal complexity of operations.
Abstract: This paper proposes a new approach for performing the secure video transcoding based on correlation preserving sorting algorithm. The idea is to perform the transcoding operation without decrypting the encrypted video content. Correlation preserving sorting algorithm is adopted here to encrypt the video data. In this method, the first frame and the frames at which the correlation coefficient between the current frame and the previous frame is less than a threshold value are transmitted through a secure channel. The video frames are sorted and the sorted frames are encoded by using any video coding standard. It is proposed to perform the transcoding operation on the sorted frames of the video. The proposed secure video transcoding system is validated through the popular video coding standard s MPEG2 and H.264. From experiments, it is p roved that the proposed system gives better compression performance and security with minimal complexity of operations.
1 citations
24 Sep 2014
TL;DR: The proposed Φ B matrix generation technique is validated with popular CS based video reconstruction techniques and it is found that the proposed method provides an improvement in the key space and security without affecting the basic features of compressive sensing.
Abstract: In this paper, a new approach for encrypting the video data through compressive sensing is proposed. Even though, Orsdemir’s cryptography key based measurement matrix (Φ B matrix) generation technique [11] provides a robust encryption method for CS framework, this scheme can not provide large key space and security. Hence the aim of this work is to improve the security and key-space of the compressive sensing paradigm without affecting the basic features of compressive sensing such as good reconstruction performance, robustness to noise characteristics and also a low encoder complexity. In order to improve the key space and security, piecewise linear chaotic map (PWLCM) based of Φ B matrix generation technique is proposed. The PWLCM is run for random number of iterations to form a random array which is used as the initial seed for generating the secure Φ B matrix. The initial & system parameter values and the number of iterations of PWLCM are kept as secret key. The proposed Φ B matrix generation technique is validated with popular CS based video reconstruction techniques and it is found that the proposed method provides an improvement in the key space and security without affecting the basic features of compressive sensing.