Secure and Private AI with PySyft
rakshit naidu (~rakshit04) |
Pysyft is a library in Python that focuses on three main techniques: Differential Privacy, Federated Learning and Secure MPC(Multi-Party Computation). This framework was built by researchers from University of Oxford and various other developers have contributed towards this amazing library. Cynthia Dwork's paper on Algorithmic Foundations of Differential Privacy gives us a great intuition on how datasets can be exploited and used for compromising the privacy of an individual. Many concepts were introduced in this breakthrough paper like Sensitivity, Randomizing the Response, Epsilon-Differential Privacy, PATE(Private Aggregation of Teacher Ensembles) and I will be sharing insights into some of the famous algorithms used for preserving the privacy of an individual in this poster.
Basics of Statistics, Number Theory, Linear Algebra, Machine Learning, Deep Learning with Pytorch.
Rakshit Naidu is a 3rd year undergraduate student from Manipal Institute of Technology. He is publishing a paper on Information Abuse in Academic Data in a Q1 Journal called the Journal of Cyber Security and this paper uses Differential Privacy and Federated Learning techniques to preserve students' data. He is the AI Head at a student project called MotoManipal which focuses on making an electric bike for MotoStudent 2020 which will be held in Spain. He is also the co-lead of the only Cryptographic Research Group at his college where he is working on Cryptanalysis on an algorithm called Keccak. He also received the registration scholarship worth $150 to the Linux Open Source Security Summit North America 2019 that was held in San Diego, California. He has also developed an Action for Google called 'Mumbai Trivia'. He is passionate about research and aims to pursue his Master's at a reputed university in the United States.