Xin Jin

Hello!

I am a double-degree master student at TU Berlin(Berlin, Germany) & KTH(Stockholm, Sweden), majoring in ICT (Information and Communication Technology) & EE(Electrical Engineering). I am pursuing my master thesis internship for federated learning’s hardware security enhancement techniques at Trustworthy & Applied Cryptography Lab.

Currently

I am currently working on secure learning. I like discussing potentially feasible methods for practical problems with nice and talented people, from diverse backgrounds, feel free to reach out! I am working on Intel SGX development for a large hybrid-secure platform project concentrating on cloud secure computation. I am looking for PhD/RA/Research Intern opportunities, working on Secure machine learning/Trustworthy AI/Computer Architecture/HPC/System security/Optimization/Applied cryptography, if you are interested in working with me, feel free to reach out!

Recent Work/Research Experience

Security Research Internship, Trustworthy and applied cryptography lab

2023 June. - Dec.

I am working as a research internship student at Trustworthy and Applied Cryptography Lab, Huawei Munich Center, Germany. I deployed machine learning’s TEEs(Trusted Execution Environment) low-level interfaces based on Intel SGX. My C/C++ programming skills have grown a lot, happy to get to know applied cryptography knowledge such as DP(differential Privacy), MPC(Multi-Party Computation) cryptography methods etc. With the opportunity entrying security research, I find the topic of how to prevent system from attacks and leakages very interesting, and would like to continue contributing on the secure computation work in the future.

Enhancing Privacy, Integrity and Security in Federated Learning: Intel SGX-based Multi-Party Computation Platform

2023 June. - 2024 Jan.

Abstract: With the booming development of large dataset collection and deep learning algorithm’s success into people’s life, Federated Learning(FL), which enables guests to train the desired effective model without sharing privacy with other clients, is raising more and more importance at various scenarios. However, certain deep learning models, such as Generative Adversarial Networks (GANs), pose privacy risks in FL, as they can inadvertently reveal sensitive information from separate devices. This poses a potential threat to users’ privacy and necessitates the development of robust privacy-preserving techniques.

Master thesis is supervised by Dr. Tianxiang Dai at lab. University supervisor: Prof. Ming Xiao and University examiner: Prof. Johan Håstad

Research Assistant, Deep learning and Image processing

2018 -2020

I worked as a research assistant for Prof. Shibai Yin on the deep learning techniques such as transformer, combining traditional computer image theories such as atmosphere scattering model for image dehazing. The useful tools are pytorch, linux shell and git.

《Image Dehazing with Uneven Illumination Prior by Dense Residual Channel Attention Network》.

A BSR-CSR dynamic sparse storage format For Sparse matrix with dense submatrix

Spring 2021

I researched and suggested a sparse matrix storage methods to accelerate special datasets training. The sparse matrix type is a large sparse matrix with sub-dense matrix, and there are many medical datasets such as dataset melanoma dataset with such feature. The Bachelor thesis was supervised by Prof. Yao Chen

See the final report: 《Storing and Computing Sparse Matrices with Dense Submatrices》.

Check out my cv page for more.

Projects

(in constructing……)

Check out my projects page for more.