Showing 1–7 of 7 results for author: Yadav, D K

Compressive Sensing Based Adaptive Defence Against Adversarial Images

Authors: Akash Kumar Gupta, Arpan Chattopadhyay, Darpan Kumar Yadav

Abstract: Herein, security of deep neural network against adversarial attack is considered. Existing compressive sensing based defence schemes assume that adversarial perturbations are usually on high frequency components, whereas recently it has been shown that low frequency perturbations are more effective. This paper proposes a novel Compressive sensing based Adaptive Defence (CAD) algorithm which combat… ▽ More Herein, security of deep neural network against adversarial attack is considered. Existing compressive sensing based defence schemes assume that adversarial perturbations are usually on high frequency components, whereas recently it has been shown that low frequency perturbations are more effective. This paper proposes a novel Compressive sensing based Adaptive Defence (CAD) algorithm which combats distortion in frequency domain instead of time domain. Unlike existing literature, the proposed CAD algorithm does not use information about the type of attack such as l0, l2, l-infinity etc. CAD algorithm uses exponential weight algorithm for exploration and exploitation to identify the type of attack, compressive sampling matching pursuit (CoSaMP) to recover the coefficients in spectral domain, and modified basis pursuit using a novel constraint for l0, l-infinity norm attack. Tight performance bounds for various recovery schemes meant for various attack types are also provided. Experimental results against five state-of-the-art white box attacks on MNIST and CIFAR-10 show that the proposed CAD algorithm achieves excellent classification accuracy and generates good quality reconstructed image with much lower computation △ Less

Submitted 11 October, 2021; originally announced October 2021.

A Two-Systems Perspective for Computational Thinking

Authors: Arvind W Kiwelekar, Swanand Navandar, Dharmendra K. Yadav

Abstract: Computational Thinking (CT) has emerged as one of the vital thinking skills in recent times, especially for Science, Technology, Engineering and Management (STEM) graduates. Educators are in search of underlying cognitive models against which CT can be analyzed and evaluated. This paper suggests adopting Kahneman's two-systems model as a framework to understand the computational thought process. K… ▽ More Computational Thinking (CT) has emerged as one of the vital thinking skills in recent times, especially for Science, Technology, Engineering and Management (STEM) graduates. Educators are in search of underlying cognitive models against which CT can be analyzed and evaluated. This paper suggests adopting Kahneman's two-systems model as a framework to understand the computational thought process. Kahneman's two-systems model postulates that human thinking happens at two levels, i.e. fast and slow thinking. This paper illustrates through examples that CT activities can be represented and analyzed using Kahneman's two-systems model. The potential benefits of adopting Kahneman's two-systems perspective are that it helps us to fix the biases that cause errors in our reasoning. Further, it also provides a set of heuristics to speed up reasoning activities. △ Less

Submitted 6 December, 2020; originally announced December 2020.

Comments: Accepted version of the paper for 12th International Conference on Intelligent Human Interaction (IHCI 2020) held from 24th to 26th November 2020 at Exco-Daegu South Korea

Structural, Elastic, Electronic and Magnetic Properties of MnNbZ (Z=As, Sb) and FeNbZ (Z=Sn, Pb) Semi-Heusler Alloys

Authors: D. K. Yadav, S. R. Bhandari, G. C. Kaphle, Madhav Prasad Ghimire

Abstract: The study of structural, electronic, magnetic, and elastic properties of new series of semi-Heusler alloys MnNbZ (Z=As, Sb) and FeNbZ (Z=Sn, Pb) has been performed by density functional theory. The magnetic phase and hence the structural stability of the alloys were considered wherein ferromagnetic state is found to stable. The half-metallic states are observed from the density of states and band… ▽ More The study of structural, electronic, magnetic, and elastic properties of new series of semi-Heusler alloys MnNbZ (Z=As, Sb) and FeNbZ (Z=Sn, Pb) has been performed by density functional theory. The magnetic phase and hence the structural stability of the alloys were considered wherein ferromagnetic state is found to stable. The half-metallic states are observed from the density of states and band structure calculations. The total magnetic moments found for all studied compounds are 1 $μ_B$/f.u., which obey Slating-Pauling rule for semi-Heusler with ferromagnetic behavior. The calculated elastic constant C$_{ij}$, cohesive energy, and formation energy confirmed that these materials are mechanically stable. Among the four system, MnNbAs is found to have the highest ductility while the remaining systems are found to be brittle in nature. These properties confirmed that among others, MnNbAs is one of the novel candidate for spintronic devices applications. △ Less

Submitted 9 September, 2020; originally announced September 2020.

Comments: 20 pages, 7 figures, Full-paper article submitted to Mat. Res. Exp

Efficient detection of adversarial images

Authors: Darpan Kumar Yadav, Kartik Mundra, Rahul Modpur, Arpan Chattopadhyay, Indra Narayan Kar

Abstract: In this paper, detection of deception attack on deep neural network (DNN) based image classification in autonomous and cyber-physical systems is considered. Several studies have shown the vulnerability of DNN to malicious deception attacks. In such attacks, some or all pixel values of an image are modified by an external attacker, so that the change is almost invisible to the human eye but signifi… ▽ More In this paper, detection of deception attack on deep neural network (DNN) based image classification in autonomous and cyber-physical systems is considered. Several studies have shown the vulnerability of DNN to malicious deception attacks. In such attacks, some or all pixel values of an image are modified by an external attacker, so that the change is almost invisible to the human eye but significant enough for a DNN-based classifier to misclassify it. This paper first proposes a novel pre-processing technique that facilitates the detection of such modified images under any DNN-based image classifier as well as the attacker model. The proposed pre-processing algorithm involves a certain combination of principal component analysis (PCA)-based decomposition of the image, and random perturbation based detection to reduce computational complexity. Next, an adaptive version of this algorithm is proposed where a random number of perturbations are chosen adaptively using a doubly-threshold policy, and the threshold values are learnt via stochastic approximation in order to minimize the expected number of perturbations subject to constraints on the false alarm and missed detection probabilities. Numerical experiments show that the proposed detection scheme outperforms a competing algorithm while achieving reasonably low computational complexity. △ Less

Submitted 9 July, 2020; originally announced July 2020.

Comments: 10 pages, 3 figures, 3 algorithms, 8 tables. Extension of the Conference paper:- Kartik Mundra, Rahul Modpur, Arpan Chattopadhyay, and Indra Narayan Kar. Adversarial image detection in cyber-physical systems. In Proceedings of the 1st ACM Workshop on Autonomous and Intelligent Mobile Systems, pages 1-5, 2020. Can be found at https://dl.acm.org/doi/abs/10.1145/3377283.3377285

Orthogonal Ramanujan Sums, its properties and Applications in Multiresolution Analysis

Authors: Devendra Kumar Yadav, Gajraj Kuldeep, S. D. Joshi

Abstract: Signal processing community has recently shown interest in Ramanujan sums which was defined by S.Ramanujan in 1918. In this paper we have proposed Orthog- onal Ramanujan Sums (ORS) based on Ramanujan sums. In this paper we present two novel application of ORS. Firstly a new representation of a finite length signal is given using ORS which is defined as Orthogonal Ramanujan Periodic Transform.Secon… ▽ More Signal processing community has recently shown interest in Ramanujan sums which was defined by S.Ramanujan in 1918. In this paper we have proposed Orthog- onal Ramanujan Sums (ORS) based on Ramanujan sums. In this paper we present two novel application of ORS. Firstly a new representation of a finite length signal is given using ORS which is defined as Orthogonal Ramanujan Periodic Transform.Secondly ORS has been applied to multiresolution analysis and it is shown that Haar transform is a spe- cial case. △ Less

Submitted 24 May, 2017; originally announced July 2017.

Ramanujan Sums as Derivatives

Authors: Devendra Kumar Yadav, Gajraj Kuldeep, S. D. Joshi

Abstract: In 1918 S. Ramanujan defined a family of trigonometric sum now known as Ramanujan sums. In the last few years, Ramanujan sums have inspired the signal processing community. In this paper, we have defined an operator termed here as Ramanujan operator. In this paper it has been proved that these operator possesses properties of first derivative and second derivative with a particular shift. Generali… ▽ More In 1918 S. Ramanujan defined a family of trigonometric sum now known as Ramanujan sums. In the last few years, Ramanujan sums have inspired the signal processing community. In this paper, we have defined an operator termed here as Ramanujan operator. In this paper it has been proved that these operator possesses properties of first derivative and second derivative with a particular shift. Generalised multiplicative property and new method of computing Ramanujan sums are also derived in terms of interpolation. △ Less

Submitted 10 November, 2016; originally announced November 2016.

A Quantitative Understanding of Human Sex Chromosomal Genes

Authors: Sk. Sarif Hassan, Pabitra Pal Choudhury, Antara Sengupta, Binayak Sahu, Rojalin Mishra, Devendra Kumar Yadav, Saswatee Panda, Dharamveer Pradhan, Shrusti Dash, Gourav Pradhan

Abstract: In the last few decades, the human allosomes are engrossed in an intensive attention among researchers. The allosomes are now already been sequenced and found there are about 2000 and 78 genes in human X and Y chromosomes respectively. The hemizygosity of the human X chromosome in males exposes recessive disease alleles, and this phenomenon has prompted decades of intensive study of X-linked disor… ▽ More In the last few decades, the human allosomes are engrossed in an intensive attention among researchers. The allosomes are now already been sequenced and found there are about 2000 and 78 genes in human X and Y chromosomes respectively. The hemizygosity of the human X chromosome in males exposes recessive disease alleles, and this phenomenon has prompted decades of intensive study of X-linked disorders. By contrast, the small size of the human Y chromosome, and its prominent long-arm heterochromatic region suggested absence of function beyond sex determination. But the present problem is to accomplish whether a given sequence of nucleotides i.e. a DNA is a Human X or Y chromosomal genes or not, without any biological experimental support. In our perspective, a proper quantitative understanding of these genes is required to justify or nullify whether a given sequence is a Human X or Y chromosomal gene. In this paper, some of the X and Y chromosomal genes have been quantified in genomic and proteomic level through Fractal Geometric and Mathematical Morphometric analysis. Using the proposed quantitative model, one can easily make probable justification or deterministic nullification whether a given sequence of nucleotides is a probable Human X or Y chromosomal gene or not, without seeking any biological experiment. Of course, a further biological experiment is essential to validate it as the probable Human X or Y chromosomal gene homologue. This study would enable Biologists to understand these genes in more quantitative manner instead of their qualitative features. △ Less

Submitted 1 December, 2013; v1 submitted 23 July, 2012; originally announced July 2012.