Computer Science and Engineering

<p>Shiv Nadar University is a comprehensive, multidisciplinary, research-focused, and student-centric University offering a full range of academic programs at the undergraduate, postgraduate and doctoral level. With state-of-the-art infrastructure, the University comprises of academic wings, sophisticated labs, international standard sports facilities, amphitheaters, auditorium, conference rooms and smart classrooms. The University&rsquo;s goal is to become internationally recognized for the quality of its research and creative endeavors and their applicability to improving quality of life, generating new insights and expanding the boundaries of human knowledge creativity. Committed to excellence in teaching, research and service, the University aims to serve the higher education needs of India and the world beyond.</p>

Shiv Nadar University

Due to the rise in automotive accidents across the globe, a cheap and reliable system is needed that can be retrofitted in any type of vehicle and can monitor road collision events. This research is aimed to develop an IoT system that uses contemporary smartphone’s intrinsic sensors to accurately report vehicle collision accidents on the road. Absolute linear acceleration (ALA) and the speed of the vehicle have been used to train and test our Support Vector Machine (SVM) based collision detection model. During the testing, the accuracy of the model was found to be very high with a MAPE (mean absolute percentage error) of 0.6\%. © 2021, Springer Nature Singapore Pte Ltd.

World University root node

(SoE) School of Engineering

Advances in Intelligent Systems and Computing

Modeling IoT Based Automotive Collision Detection System Using Support Vector Machine

Placement of items on the shelf space of retail stores signifcantly impacts the revenue of the retailer. Given the prevalence and popularity of medium-to-large-size retail stores, several research efforts have been made towards facilitating item/itemset placement in retail stores for improving retailer revenue. However, they do not consider the issue of urgency of sale of individual items. Hence, they cannot efficiently index, retrieve and place high-revenue itemsets in retail store slots in an urgency-aware manner. Our key contributions are two-fold. First, we introduce the notion of urgency for retail itemset placement. Second, we propose the urgency-aware URI index for efficiently retrieving high-revenue and urgent itemsets of different sizes. We discuss the URIP itemset placement scheme, which exploits URI for improving retailer revenue. We also conduct a performance evaluation with two real datasets to demonstrate that URIP is indeed effective in improving retailer revenue w.r.t. existing schemes. © 2021, Springer Nature Switzerland AG.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

An Urgency-Aware and Revenue-Based Itemset Placement Framework for Retail Stores

Location-based services refer to services that use location as primary input. But accessing user's location by an adversary invites issues of privacy breach. Instead of specific location coordinates, its surrounding area known as cloaking region is revealed in order to get the service. K anonymity technique of location privacy ensures that at least K-1 users should be included within a specific cloaked region. Researches have established that on combining K anonymity with the idea of including similar users together in a cloaked region provides stringent privacy (especially from background and heterogeneity attacks). This work quantifies the amount of privacy gain attained through, opting-for users with similar profiles instead of random users. The quantification is done by using KL divergence. Values of KL divergence of user profiles have been calculated for different cloaking regions containing similar and random users. Low KL divergence values depict privacy gains up to 33% for users with similar profiles. © 2021 IGI Global. All rights reserved.

International Journal of Information Security and Privacy

Entropy-based quantification of privacy attained through user profile similarity

Fraudulent Resource Consumption (FRC) attacks can be synthesized by subtly consuming metered resources of the cloud servers over a sustained period of time. The objective of the attacker in such attacks is to exploit the utility pricing model by fraudulently consuming cloud resources. This skillful over-consumption of resources results in a considerable financial burden to the client. These attacks are characterized by low-intensity HTTP requests per hour, akin to requests by legitimate users. Hence, the attack requests differ in intent but not in content, which makes FRC attacks hard to detect. In this paper, we propose P-estimation detection scheme to effectively detect these attacks. This is accomplished by training several deep learning LSTM models based on the web server logs. An estimate of attack percentage is calculated and then used to deploy the appropriate detection model. This technique takes into account the dynamic nature of websites where the popularity of web pages can change with time, by retraining and updating the detection models periodically. To the best of the authors’ knowledge, this technique outperforms all the existing FRC detection techniques with a False Negative Rate (FNR) and False Positive Rate (FPR) of 0.0059\% and 0.0\% respectively. The proposed technique is able to detect attacks as low as 2\% intensity. In addition to the detection scheme, this paper also delivers a mitigation and attribution technique to identify such attackers and block them. © 2020 Elsevier Ltd

Journal of Information Security and Applications

Detection and mitigation of fraudulent resource consumption attacks in cloud using deep learning approach

The Visual Question Answering (VQA) task combines challenges for processing data with both Visual and Linguistic processing, to answer basic ‘common sense’ questions about given images. Given an image and a question in natural language, the VQA system tries to find the correct answer to it using visual elements of the image and inference gathered from textual questions. In this survey, we cover and discuss the recent datasets released in the VQA domain dealing with various types of question-formats and robustness of the machine-learning models. Next, we discuss about new deep learning models that have shown promising results over the VQA datasets. At the end, we present and discuss some of the results computed by us over the vanilla VQA model, Stacked Attention Network and the VQA Challenge 2017 winner model. We also provide the detailed analysis along with the challenges and future research directions. © 2021, Springer Nature Singapore Pte Ltd.

Communications in Computer and Information Science

Visual Question Answering Using Deep Learning: A Survey and Performance Analysis

Machine learning is the key solution to many AI issues, but learning models rely heavily on specific training data. While a Bayesian setup can be used to incorporate some learning patterns with previous knowledge, those patterns can not access any organized world knowledge on requirements. The primary objective is to enable human-capable machines in ordinary everyday circumstances to estimate and make presumptions. In this paper we propose to respond to such common sense issues through a textual inference system with external, organized common sense graphs for explanatory inferences. The framework is based on a schematic map as a pair of questions and answers, a linked subgraph from the semantine to the symbolic space of knowledge-based external information. It displays a schematic map with a new network graphic module for information knowledge and performance with graph representations. LSTMs and graphical networks with a hierarchical attention-based direction are the basis of our model. It is flexible and understandable from the intermediate attention scores, leading to confident results. We also achieved state-of-the-art reliability on CommonsenseQA, a broad database of common sense reasoning utilizing ConceptNet as the only external tool for BERT-based models. © 2021 Institute of Physics Publishing. All rights reserved.

Fulltext

IOP Conference Series: Materials Science and Engineering

Knowledge graph augmented advanced learning models for commonsense reasoning

Retailer revenue is significantly impacted by item placement in retail stores. Notably, placement of items in the premium slots (i.e., slots with increased visibility/accessibility) improves the probability of sale w.r.t. item placement in non-premium slots. Moreover, customers often tend to buy sets of items (i.e., itemsets) instead of individual purchases. In this paper, we address the problem of maximizing retailer revenue by determining the placement of itemsets in different types of slots with varied premiumness. Our key contributions are as follows. First, we introduce the notion of premiumness of retail slots and discuss the issue of itemset placement in slots with varied premiumness. Second, we propose two efficient schemes, namely P remiumness and R evenue-based I temset P lacement (PRIP) and P remiumness and A verage R evenue-based I temset P lacement (PARIP), for placing itemsets with varying revenue in slots with varied premiumness. Third, we perform a detailed performance analysis using both real and synthetic datasets to showcase the effectiveness of our proposed schemes. We also perform a comprehensive mathematical analysis of our proposed schemes w.r.t. the complexity analysis. © 2013 IEEE.

IEEE Access

A Retail Itemset Placement Framework Based on Premiumness of Slots and Utility Mining

Manipulation of images, videos and audios using face edit apps and web services have long been in use, since decades but recent advances in deep learning has led to rising AI generated fake images and videos with swapped faces, lip synced audios and puppet masters, popularly known as Deepfakes. Generated primarily using one of the following two approaches namely, Autoencoders and Generator Adversarial Networks which rests on trained deep neural networks, deepfakes offer unprecedented challenges. The degree of realism achieved by deep learning powered deepfakes increases with increasing amounts of data i.e, fake images and videos readily available on the internet at disposal to train GANs. Deepfake algorithms create media leaving a bare margin of difference between the authentic or original source and the forged or deepfaked targets. Thus, new mechanisms and techniques to detect and filter out such deepfakes is the need of the hour.This paper exploits two powerful deep learning based CNN architectures namely, Inception-Resnet-v2 and XceptionNet for detecting the deepfakes. The proposed approach not only outshines the existing approaches in terms of efficiency and accuracy but also offers the best in terms of the given space and time complexity. © 2021 IEEE.

Proceedings - 5th International Conference on Computing Methodologies and Communication, ICCMC 2021

Demystifying deepfakes using deep learning

A working person on an average spends 1.5–2 h every day traveling either to their places of work or for other daily activities, using metros, trams, buses, and cars, as common modes of travel. Most of such commuters regularly suffer from health conditions like headache, breathless condition, drowsiness, etc. Numerous accidents have been reported due to drowsiness while driving, which may occur due to the build-up of carbon-dioxide (CO2) build in the vehicle chamber. This paper attempts to monitor, analyze, and predict air quality inside the vehicle. This work proposes a sensing system using an off-the-shelf sensor Sensordrone which is connected to an Android Smartphone using Bluetooth Low Energy. The data obtained from the proposed sensing system are then utilized to perform predictive analysis of CO2 build-up inside the vehicular chamber using Auto Regressive Integrated Moving Average (ARIMA) and Support Vector Regression (SVR). Root-Mean-Square Error for SVR and ARIMA models is 47.91 ppm and 55.32 ppm CO2, respectively, indicating that SVR outperformed ARIMA in predicting the CO2 build-up inside the vehicle. © 2021, The Author(s), under exclusive licence to Springer Nature Switzerland AG part of Springer Nature.

Journal of Reliable Intelligent Environments

Modeling vehicle indoor air quality using sensor data analytics

Transfer learning enables solving a specific task having limited data by using the pre-trained deep networks trained on large-scale datasets. Typically, while transferring the learned knowledge from source task to the target task, the last few layers are fine-tuned (re-trained) over the target dataset. However, these layers are originally designed for the source task that might not be suitable for the target task. In this paper, we introduce a mechanism for automatically tuning the Convolutional Neural Networks (CNN) for improved transfer learning. The pre-trained CNN layers are tuned with the knowledge from target data using Bayesian Optimization. First, we train the final layer of the base CNN model by replacing the number of neurons in the softmax layer with the number of classes involved in the target task. Next, the CNN is tuned automatically by observing the classification performance on the validation data (greedy criteria). To evaluate the performance of the proposed method, experiments are conducted on three benchmark datasets, e.g., CalTech-101, CalTech-256, and Stanford Dogs. The classification results obtained through the proposed AutoTune method outperforms the standard baseline transfer learning methods over the three datasets by achieving 95.92%, 86.54%, and 84.67% accuracy over CalTech-101, CalTech-256, and Stanford Dogs, respectively. The experimental results obtained in this study depict that tuning of the pre-trained CNN layers with the knowledge from the target dataset confesses better transfer learning ability. The source codes are available at https://github.com/JekyllAndHyde8999/AutoTune_CNN_TransferLearning. © 2020 Elsevier Ltd