OpenMined

Mentors: Student: Param Mirani

<p>Openmined's Syft library provides an infrastructure for computing on data you do not own and cannot see. It allows data scientists to work with data owners and provide machine learning solutions while preserving the privacy of the data. Currently, its infrastructure allows users to train models using libraries like PyTorch. While Pytorch implements many deep learning models, it doesn't implement many machine learning algorithms like SVM and Random forest. <strong>This limits users of Syft in terms of using variety of algorithms to train models on remote data.</strong> Scikit-learn provides many simple and efficient tools for predictive data analysis and is very popular and is used vastly in the Data-Scientist community. <strong>This project aims to integrate Scikit-learn into Syft's infrastructure to allow users of Syft to use all end-to-end algorithms and features which Scikit-learn provides.</strong></p>

Mentors: Student: vsquareg

<p>This project aims at firstly, investigating the performance of the current implementations against its alternatives at various levels of data transfer over the network connection for Duet, and further integrating the improvements into the PySyft stack.</p>

Mentors: Student: Aditi Verma

<p>A project to integrate NumPy, the leading package for scientific computations, into PySyft, a library for answering questions with data you cannot see.</p> <p>This project will integrate core NumPy functionalities into PySyft, such as N-dimensional array manipulation, linear algebra and other high-level mathematical functions, as well as random-number capabilities.</p> <p>While NumPy has widespread applications in the scientific domain, integrating it with PySyft will enable efficient mathematical computations to be brought to the field of Privacy-Preserving Machine Learning (PPML).</p>

Mentors: Student: Rasswanth S

<p>FALCON is the current state-of-the-art Multiparty Computation (MPC) Framework for Private-Deep Learning(PDL).MPC allows mutually distrusting parties to jointly compute a function without revealing their private inputs beyond what is revealed by the output of the computation. PDL allows multiple parties to collaboratively train a Machine Learning model without revealing their private inputs. It also encompasses secure prediction.MPC is one of the privacy advocates for PDL.</p> <p>The project aims to implement FALCON in the SyMPC library. It would greatly improve the performance of the SyMPC library for training and inference.</p>

Mentors: Student: arpitvaghela

<p>The Syft ecosystem allows one to write software that can compute over information you do not own on machines you do not have (total) control over. PySyft mainly allows <strong>direct computation on unseen data</strong> and generates static graphs of computation that can be deployed or scaled later.</p> <p>To do so PySyft creates an AST that maps function calls to their exact path and knows what to do with a node in the tree. AST allows remote execution, and to do so AST provides a local handler for the result of remote execution through pointer alongside generating all the permitted methods and attributes on a Pointer.</p> <p>The current implementation has integration for pandas’ DataFrame and Series, i.e, one can <strong>remotely execute methods and access attributes</strong> on objects on these classes using their pointers. This is to be <strong>further extended for all the classes and global functions in pandas</strong>.</p>