Who am I ?

Ph.D. Student at University of Edinburgh

I am currently a Ph.D. student in computer science at the University of Edinburgh. I am very honored to be advised by Professor Rik Sarkar. My main research interests are Differential Privacy and Machine Learning. Formerly studied my bachelor studies at the Sharif University of Technology in Applied Mathematics where I mostly focused on Differential Privacy, Coding Theory, and Information Theory under the supervision of Professor Javad Ebrahimi. Also, I had the great honor to work with Professor Marco Mondelli on Channel Coding and Machine Learning as a scientific intern at IST Austria.

Download My CV

Personal Info

  • Birthdate: 06/03/2001
  • Address: Informatics Department, University of Edinburgh, Edinburgh, United Kingdom
  • Email: s.torkamani@sms.ed.ac.uk

Research Interests

Differential Privacy

Theoritical Machine Learning

Statistical Inference

My Resume

Research Projects

Potential Privacy Risk in Machine Learning

December 2024 - Ongoing

We consider the problem of advance estimation of privacy risk of adding a data point to the training set of a machine learning model with standard optimisation methods.


Secure Data Scoring and Selection for Machine Learning Models with Zero-/Knowledge Proofs

December 2024 - April 2025

We describe secure protocols that evaluate an encrypted setup while ensuring efficient computation. The protocols are based on zero-knowledge proofs of properties of the data and model inference. They make use of a small subset of the data to perform an assessment and output a score that can be used to determine the final price of the data transaction.


Continual Observation with Differential Privacy

March 2024 - September 2024

We consider the problem of counting under continual observation and present a new generalization of the binary tree mechanism that uses a k-ary number system with negative digits to improve the privacy-accuracy trade-off.


Sparse MultiDecoder Recursive Projection Aggregation for Reed-Muller Codes

March 2022 - September 2022

Reed-Muller codes are one of the oldest families of codes. Following Dorsa Fathollahi and Professor Marco Mondelli’s previous paper, a sparse recursive projection aggregation (SRPA) decoder has been proposed, which achieves a performance that is close to the maximum likelihood decoder for short-length RM codes. In this project, we simulated an algorithm based on a neural network to lower the computational budget while keeping a performance close to that of the SRPA and RPA decoder by performing a better selection of projections in each sparsified decoder.


Algorithms and Differential Privacy via Graphs

March 2021 - February 2024

In this project, we have generalized the previous framework for designing utility-optimal differentially private (DP) mechanisms via graphs in two main directions. First, we studied heterogeneous mechanisms where the partial mechanism can have different probability distributions at the boundary. Secondly, we studied a general heterogeneous privacy setting on neighboring datasets which provides different levels of privacy for each. The problem is how to extend the mechanism, which is only defined at the selected vertices set, to other datasets in the graph in a computationally efficient and utility-optimal manner. We used the partial mechanism as a seed to optimally grow via the concept of the strongest induced DP condition. We showed that this can be done in polynomial time (in the size of the graph).


Education

Ph.D of Informatics

2024 - 2028

University of Edinburgh


B.sc in Applied Mathematics

2019 - 2023

Sharif University of Technology


Diploma in Physics and Mathematics Discipline

2013 – 2019

National Organization for the Development of Exceptional Talent


Skills

Python
Java
NumPy
PyTorch
Matlab
R

Languages

Persian
English
French
Italian

Latest News

Count on Your Elders: Laplace vs Gaussian Noise

January 2025

Paper “Count on Your Elders: Laplace vs Gaussian Noise” accepted at the FORC 2025. arXiv

Improved Counting under Continual Observation with Pure Differential Privacy

August 2024

Paper “Improved Counting under Continual Observation with Pure Differential Privacy” accepted at the TPDP 2024. arXiv

Optimal Differential Privacy via Graphs

October 2023

Paper “Optimal Differential Privacy via Graphs” accepted at the JSAIT 2023. arXiv

Heterogeneous Differential Privacy via Graphs

April 2022

Paper “Heterogeneous Differential Privacy via Graphs” accepted at the 2022 IEEE International Symposium on Information Theory (ISIT’22) arXiv

Beyond Research

Photography

My passion for photography sparked during our travels- it's just the beginning!

London, UK, October 2024

Westminister Bridge
Cooking

Cooking is my therapy. An escape to where I can spend time creating magic.

Italian Cuisine

Pasta Carbonara
Art, Art, Art

Art, in all its forms, is a remedy for our souls, whether through music, cinema, painting, or sculpture. With one life to live, I want to express my thoughts in art in every vibrant way possible.

Trying Watercolour For The First Time Creating May From "It Takes Two"! Oil Painting of Amalfi Coast, Italy