Subangkar Karmaker Shanto

Lecturer

Working as a Lecturer at Department of CSE, United International University, Bangladesh.
Courses Conducted: Object Oriented Programming, Artificial Intelligence, Bioinformatics, Algorithms, Computer Networks lab, Human Computer Interaction and Society, Technology & Engineering Ethics

Lecturer (Part-time)

Worked as a Part-time Lecturer at Department of CSE, Bangladesh University of Engineering and Technology, Bangladesh.
Courses Conducted: Structured Programming Language Lab

Research Assistant (Part-time)

Worked as a graduate research assistant under supervision of Dr. Atif Hasan Rahman  and Dr. Mohammed Eunus Ali in CSE, BUET. Multiple research projects are funded by the government of Bangladesh.

BayesBeat: Reliable Atrial Fibrillation Detection from Noisy Photoplethysmography Data

Collaborators: Sarkar Snigdha Sarathi Das, Masum Rahman, Md. Saiful Islam, Dr. Atif Hasan Rahman, Mohammad Mehedy Masud, Dr. Mohammed Eunus Ali
This work focuses on reliable prediction on Photoplethysmography (PPG) signals that noisy due to motion artifacts. We have used Bayesian Deep Learning Model implemented in Python using PyTorch framework to provide an uncertainty estimate along with the prediction. We have also implemented a Tizen Native and Background Service App in C for continuous and timed raw signal data collection from Smartwatch. Our Model beats the state of the art work both on the largest publicly available dataset and on the MIMIC-III dataset. It was the first application of Bayesian Deep Learning in this domain.
Current state: Accepted in UbiComp 2022, Published in IMWUT
Publication URL arXiv

Privacy risk of Machine Learning Models

Supervisor: Dr. Shagufta Mehnaz
In this work we are experimenting on vulnerabilities and privacy risks of Trained Classification Models introduced particularly from Model Explainability. We explore potential application of various model explanation techniques like LIME, SHAP, Counterfactual Explanation etc. for implementing model inversion attacks. We are currently working on how we can generate a more robust attack dataset capable of leaking more information about the black box model using minimal knowledge about the training dataset. We are also exploring the direction in which LIME explanations can be exploited for property inference attacks.
Current state: Ongoing

Contrastive Learning Based Approach for Patient Similarity

Supervisor: Dr. Mohammed Eunus Ali, Dr. Atif Hasan Rahman
In this work we are experimenting on contrastive learning to learn patient similarity from physiological signals, particularly Photoplethysmography (PPG) signals. Due to limited availability of dataset, we are currently conducting a case study on Atrial Fibrillation Detection from PPG signals. It is the first application of Contrastive Similarity Learning in this domain.
Current state: Ongoing
arXiv Link

Sarkar Snigdha Sarathi Das, Subangkar Karmaker Shanto, Masum Rahman, Md. Saiful Islam, Atif Rahman, Mohammad Mehedy Masud, Mohammed Eunus Ali.
BayesBeat: Reliable Atrial Fibrillation Detection from Noisy Photoplethysmography Data.
[UbiComp 2022] Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 6, 1, Article 8 (March 2022)
[pdf] [code] [arXiv]

An information system web-app developed in Django for an online restaurant hub on the perspective of Dhaka City of Bangladesh
GitHub: https://github.com/Subangkar/Foodsquare-Web-App
Docker: https://hub.docker.com/r/subangkar/foodsquare

A Deep Learning Project to generate image caption on Flickr8k Image dataset using Resnet-101 & LSTM with Attention Mechanism
GitHub: https://github.com/Subangkar/Image-Captioning-Attention-PyTorch

A Tizen Native UI app and Background Service to collect raw sensor data for analysis
GitHub (BG Service): https://github.com/Subangkar/Tizen-Sensor-Raw-Data-Saving-Service
GitHub (Native UI App): https://github.com/Subangkar/Gear-Fit-2-Sensor-Raw-Data-Sync

• ICMP Ping Spoofing: https://github.com/Subangkar/ICMP-Ping-Spoofing
• Buffer Overflow Attack using Seedlab: https://github.com/Subangkar/Buffer-Overflow-Attack-Seedlab

A project to demonstrate frequency spectrum visualization from real time audio via time domain to frequency domain conversion on Atmega32 microcontroller
GitHub: https://github.com/Subangkar/Real-Time-Audio-to-Frequency-Spectrum-Transformation-on-Atmega32