arXiv subjects

Deepak Mishra

Publications and source records attributed to Deepak Mishra.

At least 73 records · Page 4Linked to original sources

arXiv2021

Decoding Orders and Power Allocation for Untrusted NOMA: A Secrecy Perspective

The amalgamation of non-orthogonal multiple access (NOMA) and physical layer security is a significant research interest for providing spectrally-efficient secure fifth-generation networks. Observing the secrecy issue among multiplexed NOMA users, which is stemmed from successive interference cancellation based decoding at receivers, we focus on safeguarding untrusted NOMA. Considering the problem of each user's privacy from each other, the appropriate secure decoding order and power allocation (PA) for users are investigated. Specifically, a decoding order strategy is proposed which is efficient in providing positive secrecy at all NOMA users. An algorithm is also provided through which all the feasible secure decoding orders in accordance with the proposed decoding order strategy can be obtained. Further, in order to maximize the sum secrecy rate of the system, the joint solution of decoding order and PA is obtained numerically. Also, a sub-optimal decoding order solution is proposed. Lastly, numerical results present useful insights on the impact of key system parameters and demonstrate that average secrecy rate performance gain of about 27 dB is obtained by the jointly optimized solution over different relevant schemes.

cs.IT↗

arXiv2021

Domain Adaptive Egocentric Person Re-identification

Person re-identification (re-ID) in first-person (egocentric) vision is a fairly new and unexplored problem. With the increase of wearable video recording devices, egocentric data becomes readily available, and person re-identification has the potential to benefit greatly from this. However, there is a significant lack of large scale structured egocentric datasets for person re-identification, due to the poor video quality and lack of individuals in most of the recorded content. Although a lot of research has been done in person re-identification based on fixed surveillance cameras, these do not directly benefit egocentric re-ID. Machine learning models trained on the publicly available large scale re-ID datasets cannot be applied to egocentric re-ID due to the dataset bias problem. The proposed algorithm makes use of neural style transfer (NST) that incorporates a variant of Convolutional Neural Network (CNN) to utilize the benefits of both fixed camera vision and first-person vision. NST generates images having features from both egocentric datasets and fixed camera datasets, that are fed through a VGG-16 network trained on a fixed-camera dataset for feature extraction. These extracted features are then used to re-identify individuals. The fixed camera dataset Market-1501 and the first-person dataset EGO Re-ID are applied for this work and the results are on par with the present re-identification models in the egocentric domain.

cs.CV↗

arXiv2020

Target-Independent Domain Adaptation for WBC Classification using Generative Latent Search

Automating the classification of camera-obtained microscopic images of White Blood Cells (WBCs) and related cell subtypes has assumed importance since it aids the laborious manual process of review and diagnosis. Several State-Of-The-Art (SOTA) methods developed using Deep Convolutional Neural Networks suffer from the problem of domain shift - severe performance degradation when they are tested on data (target) obtained in a setting different from that of the training (source). The change in the target data might be caused by factors such as differences in camera/microscope types, lenses, lighting-conditions etc. This problem can potentially be solved using Unsupervised Domain Adaptation (UDA) techniques albeit standard algorithms presuppose the existence of a sufficient amount of unlabelled target data which is not always the case with medical images. In this paper, we propose a method for UDA that is devoid of the need for target data. Given a test image from the target data, we obtain its 'closest-clone' from the source data that is used as a proxy in the classifier. We prove the existence of such a clone given that infinite number of data points can be sampled from the source distribution. We propose a method in which a latent-variable generative model based on variational inference is used to simultaneously sample and find the 'closest-clone' from the source distribution through an optimization procedure in the latent space. We demonstrate the efficacy of the proposed method over several SOTA UDA methods for WBC classification on datasets captured using different imaging modalities under multiple settings.

cs.CV↗

arXiv2020

Effect of The Latent Structure on Clustering with GANs

Generative adversarial networks (GANs) have shown remarkable success in generation of data from natural data manifolds such as images. In several scenarios, it is desirable that generated data is well-clustered, especially when there is severe class imbalance. In this paper, we focus on the problem of clustering in generated space of GANs and uncover its relationship with the characteristics of the latent space. We derive from first principles, the necessary and sufficient conditions needed to achieve faithful clustering in the GAN framework: (i) presence of a multimodal latent space with adjustable priors, (ii) existence of a latent space inversion mechanism and (iii) imposition of the desired cluster priors on the latent space. We also identify the GAN models in the literature that partially satisfy these conditions and demonstrate the importance of all the components required, through ablative studies on multiple real world image datasets. Additionally, we describe a procedure to construct a multimodal latent space which facilitates learning of cluster priors with sparse supervision.

cs.LG↗

arXiv2020

QoS-aware Stochastic Spatial PLS Model for Analysing Secrecy Performance under Eavesdropping and Jamming

Securing wireless communication, being inherently vulnerable to eavesdropping and jamming attacks, becomes more challenging in resource-constrained networks like Internet-of-Things. Towards this, physical layer security (PLS) has gained significant attention due to its low complexity. In this paper, we address the issue of random inter-node distances in secrecy analysis and develop a comprehensive quality-of-service (QoS) aware PLS framework for the analysis of both eavesdropping and jamming capabilities of attacker. The proposed solution covers spatially stochastic deployment of legitimate nodes and attacker. We characterise the secrecy outage performance against both attacks using inter-node distance based probabilistic distribution functions. The model takes into account the practical limits arising out of underlying QoS requirements, which include the maximum distance between legitimate users driven by transmit power and receiver sensitivity. A novel concept of eavesdropping zone is introduced, and relative impact of jamming power is investigated. Closed-form expressions for asymptotic secrecy outage probability are derived offering insights into design of optimal system parameters for desired security level against the attacker's capability of both attacks. Analytical framework, validated by numerical results, establishes that the proposed solution offers potentially accurate characterisation of the PLS performance and key design perspective from point-of-view of both legitimate user and attacker.

cs.IT↗

arXiv2020

Spatio-Temporal Coverage Enhancement in Drive-By Sensing Through Utility-Aware Mobile Agent Selection

In recent years, the drive-by sensing paradigm has become increasingly popular for cost-effective monitoring of urban areas. Drive-by sensing is a form of crowdsensing wherein sensor-equipped vehicles (aka, mobile agents) are the primary data gathering agents. Enhancing the efficacy of drive-by sensing poses many challenges, an important one of which is to select non-dedicated mobile agents on which a limited number of sensors are to be mounted. This problem, which we refer to as the mobile-agent selection problem, has a significant impact on the spatio-temporal coverage of the drive-by sensing platforms and the resultant datasets. The challenge here is to achieve maximum spatiotemporal coverage while taking the relative importance levels of geographical areas into account. In this paper, we address this problem in the context of the SCOUTS project, the goal of which is to map and analyze the urban heat island phenomenon accurately. Our work makes several major technical contributions. First, we delineate a model for representing the mobile agents selection problem. This model takes into account the trajectories of the vehicles (public transportation buses in our case) and the relative importance of the urban regions, and formulates it as an optimization problem. Second, we provide two algorithms that are based upon the utility (coverage) values of mobile agents, namely, a hotspot-based algorithm that limits the search space to important sub-regions and a utility-aware genetic algorithm that enables the latter algorithm to make unbiased selections. Third, we design a highly efficient coverage redundancy minimization algorithm that, at each step, chooses the mobile agent, which provides maximal improvement to the spatio-temporal coverage. This paper reports a series of experiments on a real-world dataset from Athens, GA, USA, to demonstrate the effectiveness of the proposed approaches.

eess.SP↗

arXiv2020

Wireless Powered Protocol Exploiting Energy Harvesting During Cognitive Communications

In this letter, a novel wireless powered protocol is proposed to maximize the system throughput of an energy harvesting (EH) based cognitive radio network, while satisfying a minimum primary user rate requirement. For EH, we exploit both dedicated wireless power transfer from primary base station as well as ambient ones available due to wireless information transfer among primary and secondary users. Specifically, we prove convexity of the optimization problem and obtain semi-closed-form for globally optimal solution. Numerical results validate the analysis, and show an average performance improvement of $70\%$ over benchmark scheme for various system parameters.

cs.IT↗

arXiv2020

A Robust Pose Transformational GAN for Pose Guided Person Image Synthesis

Generating photorealistic images of human subjects in any unseen pose have crucial applications in generating a complete appearance model of the subject. However, from a computer vision perspective, this task becomes significantly challenging due to the inability of modelling the data distribution conditioned on pose. Existing works use a complicated pose transformation model with various additional features such as foreground segmentation, human body parsing etc. to achieve robustness that leads to computational overhead. In this work, we propose a simple yet effective pose transformation GAN by utilizing the Residual Learning method without any additional feature learning to generate a given human image in any arbitrary pose. Using effective data augmentation techniques and cleverly tuning the model, we achieve robustness in terms of illumination, occlusion, distortion and scale. We present a detailed study, both qualitative and quantitative, to demonstrate the superiority of our model over the existing methods on two large datasets.

cs.CV↗

arXiv2019

Optimal Green Hybrid Attacks in Secure IoT

In pursuance to understand the behavior of a potential green hybrid attacker in secure internet-of-things (IoT), this letter investigates optimal energy utilization from the attacker's viewpoint. Specifically, we propose a novel framework for optimizing the efficacy of a hybrid attacker, possessing capability to both eavesdrop and jam, while considering the underlying energy consumption. In particular, we maximize the attacker energy efficiency (AEE) in secure IoT by deriving the analytical solutions for jointly global-optimal attacking mode, eavesdropping rate, and jamming power. Numerical results, validating analytical claims, reveal that the proposed green design can offer about 45% improvement in average AEE over the relevant benchmarks.

eess.SP↗

arXiv2019

Utility-Aware Optimal Resource Allocation Protocol for UAV-Assisted Small Cells with Heterogeneous Coverage Demands

In this paper, we consider a UAV-assisted small-cell having heterogeneous users with different data rates and coverage demands. Specifically, we propose a novel utility-aware resource-allocation protocol to maximize the utility of UAV by allowing it to simultaneously serve the highest possible number of heterogeneous users with available energy resources. In this regard, first, we derive a closed-form expression for the rate-coverage probability of a user considering Rician fading to incorporate the strong line of sight (LoS) component in UAV communication. Next, since this UAV utility maximization problem is non-convex and combinatorial, to obtain the global optimal resource allocation policy we propose an iterative feasibility checking method for fixed integers ranging from lower to upper bound on the number of users that can be served by UAV. To further reduce the complexity, we formulate an equivalent problem aimed at minimizing per user energy consumption, where tight analytical relaxation on rate-coverage probability constraint is used along with semi-closed expressions for joint-optimal power and time allocation. Lastly, via detailed numerical investigation, we validate our analytical claims, present insights on the impact of key system parameters, and demonstrate that 60% more users can be served using the proposed scheme, as compared to relevant benchmarks.

cs.IT↗

arXiv2019

Variational Inference with Latent Space Quantization for Adversarial Resilience

Despite their tremendous success in modelling high-dimensional data manifolds, deep neural networks suffer from the threat of adversarial attacks - Existence of perceptually valid input-like samples obtained through careful perturbation that lead to degradation in the performance of the underlying model. Major concerns with existing defense mechanisms include non-generalizability across different attacks, models and large inference time. In this paper, we propose a generalized defense mechanism capitalizing on the expressive power of regularized latent space based generative models. We design an adversarial filter, devoid of access to classifier and adversaries, which makes it usable in tandem with any classifier. The basic idea is to learn a Lipschitz constrained mapping from the data manifold, incorporating adversarial perturbations, to a quantized latent space and re-map it to the true data manifold. Specifically, we simultaneously auto-encode the data manifold and its perturbations implicitly through the perturbations of the regularized and quantized generative latent space, realized using variational inference. We demonstrate the efficacy of the proposed formulation in providing resilience against multiple attack types (black and white box) and methods, while being almost real-time. Our experiments show that the proposed method surpasses the state-of-the-art techniques in several cases.

cs.CV↗

arXiv2019

Multi-Tag Backscattering to MIMO Reader: Channel Estimation and Throughput Fairness

Green low power networking with the least requirement of dedicated radio resources is need of the hour which has led to the upsurge of backscatter communication (BSC) technology. However, this inherent potential of BSC is challenged by hardware constraints of the underlying tags. We address this timely concern by investigating the practical efficacy of multiple-input-multiple-output (MIMO) technology in overcoming the fundamental limitations of BSC. Specifically, we first introduce a novel least-squares based channel estimation (CE) protocol for multi-tag BSC settings that takes care of both the unintended ambient reflections and the inability of tags in performing estimation by themselves. Then using it, a nontrivial low-complexity algorithm is proposed to obtain the optimal transceiver designs for the multiantenna reader to maximize the minimum value of the lower-bounded backscattered throughput among the single-antenna semi-passive tags. Additional analytical insights on both individually and jointly-optimal precoding vector and detector matrix at the reader are provided by exploring the asymptotically-optimal transceiver designs. Lastly detailed numerical investigation is carried out to validate the theoretical results and quantify the practically realizable throughput fairness. Specifically, more than seven-fold increase in the common-backscattered-throughput among tags as achieved by the proposed designs over the relevant benchmarks corroborates their practical significance.

cs.IT↗

arXiv2019

Unsupervised Anomalous Trajectory Detection for Crowded Scenes

We present an improved clustering based, unsupervised anomalous trajectory detection algorithm for crowded scenes. The proposed work is based on four major steps, namely, extraction of trajectories from crowded scene video, extraction of several features from these trajectories, independent mean-shift clustering and anomaly detection. First, the trajectories of all moving objects in a crowd are extracted using a multi feature video object tracker. These trajectories are then transformed into a set of feature spaces. Mean shift clustering is applied on these feature matrices to obtain distinct clusters, while a Shannon Entropy based anomaly detector identifies corresponding anomalies. In the final step, a voting mechanism identifies the trajectories that exhibit anomalous characteristics. The algorithm is tested on crowd scene videos from datasets. The videos represent various possible crowd scenes with different motion patterns and the method performs well to detect the expected anomalous trajectories from the scene.

cs.CV↗

arXiv2019

Optimal Least-Squares Estimator and Precoder for Energy Beamforming over IQ-Impaired Channels

Usage of low-cost hardware in large antenna arrays and low-power wireless devices in Internet-of-Things (IoT) has led to the degradation of practical beamforming gains due to the underlying hardware impairments like in-phase-and-quadrature-phase imbalance (IQI). To address this timely concern, we present a new nontrivial closed-form expression for the globally-optimal least-squares estimator (LSE) for the IQI-influenced channel between a multiantenna transmitter and single-antenna IoT device. Thereafter, to maximize the realistic transmit beamforming gains, a novel precoder design is derived that accounts for the underlying IQI for maximizing received power in both single and multiuser settings. Lastly, the simulation results, demonstrating a significant -8dB improvement in the mean-squared error of the proposed LSE over existing benchmarks, show that the optimal precoder designing is more critical than accurately estimating IQI-impaired channels. Also, the proposed jointly-optimal LSE and beamformer outperforms the existing designs by providing 24% enhancement in the mean signal power received under IQI.

eess.SP↗

arXiv2019

Learning Rotation Adaptive Correlation Filters in Robust Visual Object Tracking

Visual object tracking is one of the major challenges in the field of computer vision. Correlation Filter (CF) trackers are one of the most widely used categories in tracking. Though numerous tracking algorithms based on CFs are available today, most of them fail to efficiently detect the object in an unconstrained environment with dynamically changing object appearance. In order to tackle such challenges, the existing strategies often rely on a particular set of algorithms. Here, we propose a robust framework that offers the provision to incorporate illumination and rotation invariance in the standard Discriminative Correlation Filter (DCF) formulation. We also supervise the detection stage of DCF trackers by eliminating false positives in the convolution response map. Further, we demonstrate the impact of displacement consistency on CF trackers. The generality and efficiency of the proposed framework is illustrated by integrating our contributions into two state-of-the-art CF trackers: SRDCF and ECO. As per the comprehensive experiments on the VOT2016 dataset, our top trackers show substantial improvement of 14.7% and 6.41% in robustness, 11.4% and 1.71% in Average Expected Overlap (AEO) over the baseline SRDCF and ECO, respectively.

cs.CV↗

arXiv2019

Sum Throughput Maximization in Multi-Tag Backscattering to Multiantenna Reader

Backscatter communication (BSC) is being realized as the core technology for pervasive sustainable Internet-of-Things applications. However, owing to the resource-limitations of passive tags, the efficient usage of multiple antennas at the reader is essential for both downlink excitation and uplink detection. This work targets at maximizing the achievable sum-backscattered-throughput by jointly optimizing the transceiver (TRX) design at the reader and backscattering coefficients (BC) at the tags. Since, this joint problem is nonconvex, we first present individually-optimal designs for the TRX and BC. We show that with precoder and {combiner} designs at the reader respectively targeting downlink energy beamforming and uplink Wiener filtering operations, the BC optimization at tags can be reduced to a binary power control problem. Next, the asymptotically-optimal joint-TRX-BC designs are proposed for both low and high signal-to-noise-ratio regimes. Based on these developments, an iterative low-complexity algorithm is proposed to yield an efficient jointly-suboptimal design. Thereafter, we discuss the practical utility of the proposed designs to other application settings like wireless powered communication networks and BSC with imperfect channel state information. Lastly, selected numerical results, validating the analysis and shedding novel insights, demonstrate that the proposed designs can yield significant enhancement in the sum-backscattered throughput over existing benchmarks.

cs.IT↗

arXiv2019

Mode matching in GANs through latent space learning and inversion

Generative adversarial networks (GANs) have shown remarkable success in generation of unstructured data, such as, natural images. However, discovery and separation of modes in the generated space, essential for several tasks beyond naive data generation, is still a challenge. In this paper, we address the problem of imposing desired modal properties on the generated space using a latent distribution, engineered in accordance with the modal properties of the true data distribution. This is achieved by training a latent space inversion network in tandem with the generative network using a divergence loss. The latent space is made to follow a continuous multimodal distribution generated by reparameterization of a pair of continuous and discrete random variables. In addition, the modal priors of the latent distribution are learned to match with the true data distribution using minimal-supervision with negligible increment in number of learnable parameters. We validate our method on multiple tasks such as mode separation, conditional generation, and attribute discovery on multiple real world image datasets and demonstrate its efficacy over other state-of-the-art methods.

cs.CV↗

arXiv2019

Optimal Channel Estimation for Hybrid Energy Beamforming under Phase Shifter Impairments

Smart multiantenna wireless power transmission can enable perpetual operation of energy harvesting (EH) nodes in the internet-of-things. Moreover, to overcome the increased hardware cost and space constraints associated with having large antenna arrays at the radio frequency (RF) energy source, the hybrid energy beamforming (EBF) architecture with single RF chain can be adopted. Using the recently proposed hybrid EBF architecture modeling the practical analog phase shifter impairments (API), we derive the optimal least-squares estimator for the energy source to EH user channel. Next, the average harvested power at the user is derived while considering the nonlinear RF EH model and a tight analytical approximation for it is also presented by exploring the practical limits on the API. Using these developments, the jointly global optimal transmit power and time allocation for channel estimation (CE) and EBF phases, that maximizes the average energy stored at the EH user is derived in closed form. Numerical results validate the proposed analysis and present nontrivial design insights on the impact of API and CE errors on the achievable EBF performance. It is shown that the optimized hybrid EBF protocol with joint resource allocation yields an average performance improvement of $37\%$ over benchmark fixed allocation scheme.

cs.IT↗