Welcome all!!. :)

This page is not routinely maintained, apologies!. For details of publications, please check my linked Google scholar account. A Machine Learning enthusiast, overwhelmed by Quantum Theory, Black Holes and Human Psychology.

Selected Publications

This paper addresses the problem of pruning the gigantic social network retaining desirable properties, develops a criterion to form a final contributor graph with potential vertices and demonstrates the quality of a good trip design. Our model addresses issues concerning optimization of revenue for Online Travel Agency (OTA) while taking care of customers requirements as well. Our model discovers co-occurrence relationships among activities performed by (or recorded about) specific individuals or groups and correlate users with similar interests. The problem of finding minimum potential vertex set is computationally similar to standard NP-hard problems in graph theory like finding vertex cover, max-clique, etc. We propose a greedy approach to developing heuristic for final pruning and focus our model to provide a satisfiable good-enough solution - feasible to compute in polynomial time. We present a comprehensive summary of literature survey in domains of influence maximization, recommendation systems, etc in the context of marketing - with the emphasis on context-specific influence propagation. The results show the efficiency of our approach and substantiates our claim that this approach forms a conglomerate of social groups leading to a good trip design. Proposed model is flexible, generic and adaptable to address similar data-driven user-oriented problems.
In TKDE, 2018
Project

Since a single fiber carries a huge amount of data in optical WDM networks, a fiber cut even for a brief period is a very serious event. Designing schemes to prevent disruption of user traffic and recovery techniques from failures is thus an important area of research. Since a single fiber cut is the most common type of fault, in this paper we address the problem of protecting all-optical WDM mesh networks from single link failure. Our proposed online protection scheme is an improvement over an existing approach and is not only cost-efficient in terms of network resource consumption but can also provide quick recovery from a link failure. We first provide an ILP formulation for the problem and then propose a heuristic solution iStreams that can provide near-optimal solution in polynomial time. Performance comparisons with some well-known schemes of protection show that our heuristic algorithm can be a better choice for conserving resource while providing quick recovery from a link failure.
In PNC, Springer, Vol.35, No.1, (Impact Factor- 1.210, Indexed in - Science Citation Index (SCI), Science Citation Index Expanded (SciSearch), SCOPUS, EBSCO, INSPEC and others), 2018
PDF Project

Recent Publications

. Affinity Analysis in Social Networks for Online Travel Agency: User oriented data-driven approach.. In TKDE, 2018.

Project

. An online cost-efficient protection scheme for quick recovery in all-optical WDM mesh networks.. In PNC, Springer, Vol.35, No.1, (Impact Factor- 1.210, Indexed in - Science Citation Index (SCI), Science Citation Index Expanded (SciSearch), SCOPUS, EBSCO, INSPEC and others), 2018.

PDF Project

. Security Vulnerabilities of Unmanned Aerial Vehicle and Countermeasures: An Experimental Study.. In VLSID (To appear in IEEE Xplore), 2018.

Project

. An offline scheme for reducing cost of protection in all-optical WDM mesh networks with fast recovery. In IJPED, (Indexed in - SCOPUS (Elsevier), Emerging Sources Citation Index (Thomson Reuters), Web of Science (Thomson Reuters), ACM Digital Library, DBLP, EBSCO and others), 2017.

PDF Project

. SEMRSA: Spectrum Efficient Modulation- aware Dynamic Routing & Spectrum Allocation in Elastic Optical Network.. In ICCMS, published by ACM USA, 2017.

PDF

Recent & Upcoming Talks

IEEE VLSID 2018
Jan 10, 2018 3:45 PM
Academic Conference/Symposium
Project

Recent Posts

Projects

An Online Cost-Efficient Protection Scheme for Quick Recovery in All-Optical WDM Mesh Networks

Designing path-protection based scheme to prevent traffic disruptions from single-link failures. Developing and implementing new algorithms; an improvement over existing algorithms like streams, gives more savings in terms of network resources in dynamic environment. Heuristic gave near-optimal results for networks like USANET, NSFNET.

An offline scheme for reducing cost of protection in all-optical WDM mesh networks with fast recovery

Designing schemes to prevent disruption of user traffic from failures in offline mode. Developing and implementing new algorithms; an improvement over existing algorithms like gstream in static environment. Formulation of Integer Linear Program as a mathematical proof of optimal solution.

Developing Robust Machine Learning for deep learning applications

Study of adversarial training, threat models, poisoning, inference and evasion attacks including membership inference, generative adversarial attack (GAN) and its applications in fraud detection, malware classification, etc. Survey of existing attack models, countermeasures and its limitations, finally developing a more robust learning system.

Security Analysis of Unmanned Aerial Vehicles

Analyzed vulnerabilities of drones under study (namely DJI Phantom and Parrot Bebop), performed attacks, programmed some hijacking methods and proposed countermeasures. GPS spoofing based attacks like height hack, force-landing were successfully performed at Nanyang Technological University, Singapore.

Teaching

I led several sessions for juniors through the Robotics club, organized lectures on competitive coding. I was selected as the head of Turing Chapter of the Coding club at our esteemed university and worked hard to make seemingly tough and humongous domain of machine learning tractable. Also, I was an intern at a company GradeUp as content developer, I was assigned to set Q&A and mock tests on core CS subjects.

Tags

Contact