Yoana G. Gyurova
MIT EECS - Qualcomm Undergraduate Research and Innovation Scholar
Software solutions for unreliable hardware
Saman P. Amarasinghe
With each new generation hardware, we are gaining speed but getting more computation errors. Correcting errors on hardware level becomes extremely costly and we need ways to detect and fix these errors on software level, using fast nondeterministic checkers. The main goal of my research project is to design, implement, and evaluate these fast, and reliable program checkers for programs executed on untrusted or unreliable systems. Examples to be evaluated include checkers and extensions aimed at classical algorithms in sorting, searching, graph and number theory, machine learning, cryptography. Sometimes, extended versions of classical algorithms (such as the GCD algorithm) are easier to verify than the regular version and part of the project is finding out for which algorithms this is true.
I have been interested in working on more mathematical projects since high school where I participated in many math competitions. At MIT I took several classes that gave me stronger background in design and analysis of algorithms, as well as in number theory and cryptography and this is the main knowledge that I am going to need and use for the current project.