> Uncertainty Quantification for feature selection using Multi-Armed Bandits Feature selection for datasets with large predictors but few data points using multi armed bandits. Applying this approach to healthcare data to quantify the difference in uncertainty with this approach

> Scheduling of neural network parameters Scheduling of learning rate, batch size and other hyperparameters of neural networks for solving differential equations using reinforcement learning

> Reduction of cost of information in analysis of variable phenomena Applying reinforcement learning to classify and prioritise variable objects in order to take informed decisions regarding allocation of limited spectroscopic resources among other resources.

> Intrusion Detection System using Sensor Fusion Networks To capitalise on the large ubiquitous computing power, multiple IDSs are deployed in the same network. A neural network learns the IDSs weights for preferential attack detection. A connected sensor fusion unit makes an appropriate decision.

> Smart glasses for the blind Glasses detect all objects around it along with their distance and speed. Based on this, it actively navigates the blind, thus obviating the use of a walking stick. For a better experience it describe everything around them.

> Malware Classification as an application of social network analytics and community detection Community detection is applied to system call graphs and features are then extracted from the communities using social network properties such as the centrality measures (degree, betweenness, eigenvector, etc.)

> Comparison of Machine Learning techniques for malware detection and classification A comparison of different machine learning techniques for the detection and classification of malware variants of known malware families by correlating the static features from executables and exploiting the bloom filter data structure

> Hybrid Behavioural Features for Chrun Prediction in Mobile Telecomm Networks with Data Constriants Features based on social influence and changes in both - call usage patterns and social groups - are extracted to predict churn by underlining the importance of as social ties to better understand the behaviour of customers

> Botnet Prediction The aim is to predict which system in a network of user systems could be part of a botnet/botmaster target using predictive functions and community detection on network traffic between interconnected systems.

> Quantum Machine Learning to mitigate machine learning adoption barriers A research survey of quantum machine learning algorithms (theory & implmentational working) along with their applications to overcome issues like large data processing, training overhead. The mathematics governing quantum annealing, quantum topological algorithms and quantum NN classification throughly studied.

> Analysis of Similarity Measures for Collaborative Filtering Recommendation Implemented a user-based Collaborative Filtering (CF) mechanism to determine which similarity index (such as Manhattan distance, Pearson Correlation, etc) is ideal for data of the different forms.

> Forensic analysis of security attacks A comprehensive survey of the types of security attacks, mode of occurrence and efficient methods to analyze them. Put forward an efficient and scalable algorithm to prevent a Denial of Service Attack

> Security Analysis of AES algorithm This project involved the analysis of the design implementation of the asymmetric encryption algorithm & key security features with its complexity. Also, verified the randomness & entropy to check for the redundancy in the key generation.

> Geographic topology extrapolation using Generative Adversarial Networks Generative Adversarial Network is used to find probable topological hot-spots in urban cities for new lakes to be dugout to increase groundwater & reduce the chance of flooding of low-lying areas

> Certification Authority Management System A software project funded by Ministry of Electronics and Information Technology, Govt. of India that aids certification authorities in the validation & certification of vendor’s cryptographic modules and their underlying cryptographic algorithms. All NIST approved algorithms and modules in FIPS 104-2 are incorporated.

> Lane detection and switching in self driving cars Simulation of cars that can drive in a diverse set of virtual environment without a human operator using Convolutional Neural Networks. It is based on lane detection & mapping of steering angle with the surrounding environment.

> Electoral Database Efficient database with interactive UI and automated text summarization that tracks electoral campaigns, polling information along with candidate comparison and authentication service with double layer encryption.

> Sign Language recognition A system using which one can easily communicate with deaf people by converting American Sign Language (ASL) into text by creating a vision-based system to identify and classify fingerspelled letters of the ASL.

> Fake news detection using LSTMs A corpus of labelled news articles is used to build a Long Short-term Memory model that generates new sentences that mimic the style of the writers. This renders class imbalance is inconsequential.

> Vulnerability assessment and penetration testing tool hybrid model

> Captcha Reader