Computer Science and Engineering

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.

Shiv Nadar University

The popularity of social networks make them most efficient to integrate into the Emergency Management process. Posts on social networking sites can help people by ensuring timely detection of an emergency. Often during the situations of a natural disaster, there is an information chasm created between the affected and the unaffected area that further compounds the confusion and chaos. In this paper, we examine the various challenges that exist while attempting to integrate social networks and Emergency Management and trace the state-of-art techniques that exist in various domains that come together for this Emergency Management system. We propose an end-to-end framework that takes public posts from social networking sites and converts it into a structured format that makes the information actionable. A summarization technique may be applied to the acquired information post mining of social media feed to convert everything into a text message that can be released into various social platforms. To increase the reach of this post and to warrant better public participation in the crisis in a timely manner, we apply influence maximization techniques and monitor the diffusion process of this generated post through a diffusion modelling technique that we propose. We conduct experiments to analyze the performance of this model and of the influence maximization process and conclude with an analysis of the experiments and the observed results and list out improvements that we intend to incorporate in future versions of this work. © 2019 IEEE.

(SoE) School of Engineering

World University root node

IEEE Region 10 Annual International Conference, Proceedings/TENCON

Emergency Management using Social Networks

Location Based Social Networks (LBSNs) like Twitter, Foursquare or Instagram are a very good source of extracting human generated data in the form of check-ins, location and social relationships among users. Location data bridges the gap between the physical and digital worlds and enables a deeper understanding of users preference and behaviour. There are many underlying patterns in this type of dataset of human mobility which are utilised for applications like recommender systems. The current approaches involve extracting data from user-item rating, GPS trajectories, or other forms of data, whereas we focus on an integrated model considering factors like user's interest, social influence, time and proximity. There is metadata associated with this dataset which tells us about the whereabouts of the user, with emphasis on types of places. This paper proposes an integrated location recommendation model that considers users' interests, their friends influences, time and seasonality factors, and users' willingness to visit distant locations. We integrate all these parameters to generate a ranked list of locations which will be recommended to the user. Experiments are performed on a real-world dataset which show that our proposed model is effective in the stated conditions. © 2019 IEEE.

Influence-Time-Proximity Driven Locations Recommendation Model: An Integrated Approach

Facial expression recognition in videos is an active area of research in computer vision. However, fake facial expressions are difficult to be recognized even by humans. On the other hand, facial micro-expressions generally represent the actual emotion of a person, as it is a spontaneous reaction expressed through human face. Despite of a few attempts made for recognizing micro-expressions, still the problem is far from being a solved problem, which is depicted by the poor rate of accuracy shown by the state-of-the-art methods. A few CNN based approaches are found in the literature to recognize micro-facial expressions from still images. Whereas, a spontaneous microexpression video contains multiple frames that have to be processed together to encode both spatial and temporal information. This paper proposes two 3D-CNN methods: MicroExpSTCNN and MicroExpFuseNet, for spontaneous facial micro-expression recognition by exploiting the spatiotemporal information in CNN framework. The MicroExpSTCNN considers the full spatial information, whereas the MicroExpFuseNet is based on the 3D-CNN feature fusion of the eyes and mouth regions. The experiments are performed over CAS(ME)2 and SMIC microb expression databases. The proposed MicroExpSTCNN model outperforms the state-of-the-art methods. © 2019 IEEE.

Proceedings of the International Joint Conference on Neural Networks

Spontaneous Facial Micro-Expression Recognition using 3D Spatiotemporal Convolutional Neural Networks

The hashtag recommendation systems on Twitter have largely focused on analyzing the text content of tweets. In this work, we modify the state-of-the-art existing natural language processing (NLP) technique and deeply ingrain socio-temporal techniques into the overall process, to model a novel hashtag recommendation system. The social aspect of the system aims to make use of the hashtags generated by familiar individuals possess, as well as, the hashtags used by the individual at the past (profile). The temporal aspect aims to age the tweets, thereby ensuring that the more recent hashtags receive higher weights in the process of recommendation. The NLP technique is modified to offer an initial score based upon text embedding of hashtags, and a socio-temporal function and a burst function are applied to generate a final relevance score for hashtags towards a given tweet. The hashtags with top-K relevance scores are recommended to the user. © Springer Nature Switzerland AG 2019.

Studies in Computational Intelligence

A socio-temporal hashtag recommendation system for twitter

This work aims to collect and analyze air quality information in and around regions of Delhi using an IoT system which is placed in a moving vehicle. Hourly air quality data has been used to forecast exceedance probabilities and return periods for air pollution extremes. The exceedance probabilities and return periods obtained using Extreme Value Distribution (EVD) have been compared to actual occurrences of extreme events.The results indicate that the EVD based air quality model is best suited for forecasting air quality in this high pollutant concentration region. © 2019 IEEE.

SAS 2019 - 2019 IEEE Sensors Applications Symposium, Conference Proceedings

IoT System based Forecasting and Modeling Exceedance Probability and Return Period of Air Quality using Extreme Value Distribution

As humans spend around 90\% of their time indoors, Indoor Air Quality (IAQ) is a subject of major concern for the physical and mental well-being of humans. According to the United States Environment Protection Agency (US EPA), even in centrally air-conditioned buildings, indoor air is much more polluted than outdoor air, mainly due to changes in occupancy patterns, old or ill-maintained ventilation systems and dust. Therefore, it becomes important to measure and analyze IAQ. In this work, an end to end IoT system has been developed to sense and analyze indoor environmental parameters: temperature, humidity and carbon dioxide. These sensed values are then used to compute Predicted Mean Vote (PMV) and Ventilation Rate (VR), which are further combined to develop an indicator of the current air quality and thermal comfort index, the Air Quality and Comfort Indicator (AQCI), using fuzzy inference. © 2019 IEEE.

International Symposium on Advanced Networks and Telecommunication Systems, ANTS

AQCI: An IoT Based Air Quality and Thermal Comfort Model using Fuzzy Inference

An IoT end to end system has been developed in this work to collect relative humidity (RH), CO2 concentration and occupant count of a University classroom. The RH and CO2 data has been used to compute estimates of student occupancy using regression based estimation models. Multiple linear and quantile regression models have been explored for occupancy estimation by using RH, CO2, and both RH as well as CO2 concentration respectively. The estimation performance of these models has been compared by using mean absolute percentage error. The quantile regression based models have been found to be the most accurate with a mean absolute percentage error of 2.47\%. © 2019 IEEE.

2019 IEEE 16th India Council International Conference, INDICON 2019 - Symposium Proceedings

IoT-based occupancy estimation models for indoor non-residential environments

The problem of Scene flow estimation in depth videos has been attracting attention of researchers of machine vision, due to its potential application in various areas of robotics. The conventional scene flow estimation methods are difficult to use in real-time applications due to their long computational overhead. We propose a conditional adversarial network SceneFlowGAN for scene flow estimation. The proposed SceneFlowGAN uses loss function at two ends: both the generator and the discriminator. The proposed network is a first attempt to estimate scene flow using generative adversarial networks, and is able to estimate both the optical flow and disparity from the input stereo images simultaneously. The proposed method is experimented on a huge RGB-D benchmark sceneflow estimation dataset. © 2019 IEEE.

2019 28th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2019

A Conditional Adversarial Network for Scene Flow Estimation

Breach in password databases has been a frequent phenomena in the software industry. Often these breaches go undetected for years. Sometimes, even the companies involved are not aware of the breach. Even after they are detected, publicizing such attacks might not always be in the best interest of the companies. This calls for a strong breach detection mechanism. Juels et al. (in ACM-CCS 2013) suggest a method called ‘Honeywords’, for detecting password database breaches. Their idea is to generate multiple fake passwords, called honeywords and store them along with the real password. Any login attempt with honeywords is identified as a compromise of the password database, since legitimate users are not expected to know the honeywords corresponding to their passwords. The key components of their idea are (i) generation of honeywords, (ii) typo-safety measures for preventing false alarms, (iii) alarm policy upon detection, and (iv) testing robustness of the system against various attacks. In this work, we analyze the limitations of existing honeyword generation techniques. We propose a new attack model called ‘Multiple System Intersection attack considering Input’. We show that the ‘Paired Distance Protocol’ proposed by Chakraborty et al., is not secure in this attack model. We also propose new and more practical honeyword generation techniques and call them the ‘evolving-password model’, the ‘user-profile model’, and the ‘append-secret model’. These techniques achieve ‘approximate flatness’, implying that the honeywords generated using these techniques are indistinguishable from passwords with high probability. Our proposed techniques overcome most of the risks and limitations associated with existing techniques. We prove flatness of our ‘evolving-password model’ technique through experimental analysis. We provide a comparison of our proposed models with the existing ones under various attack models to justify our claims.

Fulltext

IEEE Transactions on Dependable and Secure Computing

Generation of Secure and Reliable Honeywords, Preventing False Detection

Efficient and precise classification of histological cell nuclei is of utmost importance due to its potential applications in the field of medical image analysis. It would facilitate the medical practitioners to better understand and explore various factors for cancer treatment. The classification of histological cell nuclei is a challenging task due to the cellular heterogeneity. This paper proposes an efficient Convolutional Neural Network (CNN) based architecture for classification of histological routine colon cancer nuclei named as RCCNet. The main objective of this network is to keep the CNN model as simple as possible. The proposed RCCNet model consists of 1, 512, 868 learnable parameters which are significantly less compared to the popular CNN models such as AlexNet, CIFAR-VGG, GoogLeNet, and WRN. The experiments are conducted over publicly available routine colon cancer histological dataset 'CRCHistoPhenotypes'. The results of the proposed RCCNet model are compared with five state-of-the-art CNN models in terms of the accuracy, weighted average F1 score and training time. The proposed method has achieved a classification accuracy of 80.61% and 0.7887 weighted average F1 score. The proposed RCCNet is more efficient and generalized in terms of the training time and data over-fitting, respectively. © 2018 IEEE.