Rohit Pai

I am a Senior Analyst in the Cryptography Engineering and Solutions team at Deutsche Bank.

For the past 2 years, I have contributed to the bank’s Post Quantum Cryptographic Transformation by assisting in the creation of the Quantum Risk Assessment Framework and developing various custom tools and packages to secure internal applications and networks.

My research interests include all aspects of cryptography, with a particular focus on Post Quantum Cryptography. Specifically, I aim to deepen my understanding of lattices and contribute to obtaining performance gains without hampering security.

Previously, I completed my BTech in Computer Engineering from Sardar Patel Institute of Technology (2022). In my capstone project, I developed a robust deep-learning-based malware detection system under the guidance of Dr. Preetida V-Jani.

In my spare time, I enjoy watching basketball (Go Lakers!), travelling, and exploring new food and cuisines with friends and family.

news

Oct 01, 2023	Promoted to Senior Analyst at Deutsche Bank
Jul 18, 2022	Thrilled to join Deutsche Bank as an Analyst in the Cryptography Engineering team!
Jan 03, 2022	Excited to join SafexPay as a Product Management Intern, contributing to the dynamic Risk Management and International Projects teams!
Dec 31, 2021	Successfuly defended my Bachelor’s Major Project!

selected publications

ICDSAI

Strengthening Deep-Learning-Based Malware Detection Models Against Adversarial Attacks

R. Pai, M. Purohit, and P. Vinayakray

In Advances in Data Science and Artificial Intelligence , 2022

Abs HTML PDF

Detecting malware is no easy feat as there are many families with exponentially more variants. Much research has been carried out into deep-learning-based malware detection models due to their ability to handle large volumes of data and learn complex features. However, classical deep learning models are unable to adapt to malware variants and specially crafted adversarial samples. Moreover, these models have been trained on older and much smaller datasets. Therefore, there is a need for new models capable of withstanding adversarial attacks while providing accurate results. Hence, we propose a model trained on the latest dataset, capable of handling various malicious adversarial samples while providing a low false-positive rate and high precision, recall, and F1 score.
ICAC3

Vigila: Application for General Safety

R. Pai, A. Naik, H. Sandesara, and S. Dholay

In 2021 International Conference on Advances in Computing, Communication, and Control (ICAC3) , 2021

Abs HTML PDF

Most safety applications already existing in the market send SOS messages to the victim’s saved contacts and the police, which could prove to be slow in situations which demand urgency. We propose a system to counter this and to respond to such emergencies. We store real-time location of our users and when anyone presses the SOS button, all users in a 100 meter radius, around that user, are notified about the emergency along with the coordinates of the victim. Our application also maintains a comprehensive first-aid and emergency response guide which the users can peruse so they can be more helpful in such situations.