Research

Listed are the various projects that I have done as part of internships and projects.

AUDITING INDIAN ELECTIONS

Vishal Mohanty, Christopher Culnane, Philip B. Stark., Vanessa Teague

I worked on this project with Prof. Vanessa Teague at the University of Melbourne during the summer of 2017 as part of a Summer Exchange Program. We designed methods for post-verification of election results for the Indian General elections. Risk-limiting audits are a way of verifying the outcome of an election with arbitrarily high confidence by just manually verifying a few ballots. We also developed a prototype of tools to compute the risk limit for various constituencies which are available at this link. Paper also available on arxiv.

REVISITING AES SBox COMPOSITE FIELD IMPLEMENTATIONS FOR FPGAs

Aditya Pradeep, Vishal Mohanty, Adarsh Muthuveeru Subramaniam, Chester Rebeiro

This is a project I worked in 2018 along with Prof. Chester Reberio and Aditya Pradeep. AES S-box is vulnerable to a variety of side-channel attacks such as power analysis. Threshold implementation prevents such side-channel attacks by sharing the various input, intermediary and output variables. We found out the most efficient implementation of the AES S-box which has the composite-field implementation coupled with the threshold implementation on FPGA.

EFFECT OF NETWORK DELAYS ON BITCOIN SELFISH MINING (Thesis)

Prof. Pandurangan Chandrasekaran, CSE Dept., IIT Madras

Bitcoin Selfish Mining is an attack that compromises the Bitcoin network by distorting the reward protocol of the mining. Generally it is considered that the Bitcoin network is secure as long as the majority of the miners in the network are honest. Selfish mining distorts that fact. As part of my thesis, I explored what effect the network propagation delay of the messages broadcast between the miners has on the effectiveness of the selfish mining attack. Interestingly, network delays affects selfish mining in a negative way and could pave way for a defence against the attack. Read on to find out more.

K-MACHINE ALGORITHMS FOR CLUSTERING PROBLEMS

Prof. John Augustine, CSE IIT Madras

k-Machine algorithms are a new way of thinking about implementation and analysis of the algorithms commonly studied in the RAM model. k interconnected computers talk to each other by sending messages via links that have a certain threshold on the amount of information that can be sent across each time. This can be thought of as a type of parallel processing with message-passing. In this setting we looked at four clustering algorithms which become very efficient- k-Means, Fuzzy-C-Means, Hierarchical and k-Means approximation. The report for this is available at this link.

HYPERSPECTRAL IMAGE PROCESSING

Prof. Aurobindo Routray, Dept. of Electrical Engineering, IIT Kharagpur

A normal RGB image has 3 layers - Red, Green and Blue. A hyperspectral image on the other hand has a large number of layers, about 200 capturing various frequencies. This helps to plot a characteristic of a certain location of an image which determines the element present at that location. Drones flying over a city equipped with these hyperspectral cameras can give information about the hazardous wastes accumulating and report to the health department of the city council. I was fortunate to develop a GUI for a open-source hyperspectral image processing application using python in my first summer in 2016.