Secure learning
Protecting deployed models against extraction while preserving the predictions legitimate users rely on.
Model extraction · Antidistillation
Explore secure learningRIT / KAS Lab · Rochester, NY
Khawaja Abaid Ullah.
PhD Student · Machine Learning & Autonomous Systems
Fulbright ScholarFormer AWS AI/ML Intern
I work on machine learning security and autonomous systems—from protecting deployed models against extraction to developing coordinated multi-agent robots for search-and-rescue.
Rochester Institute of Technology
Khojasteh Autonomous Systems Laboratory
Security. Coordination. Autonomy.
01 / RESEARCH
Secure models. Coordinated robots.
I work on model extraction defenses and multi-agent autonomy, with the goal of coordinated robots for search-and-rescue.
Autonomous multi-agent systems
Developing coordinated teams of robots for search-and-rescue in disaster-struck regions.
Reinforcement learning · Coordination
Explore multi-agent systemsMachine reasoning & optimization
Current directionExploring language-model agents for search and decision-making under partial information.
LLMs · Agents · Optimization
Explore current direction02 / SELECTED WORK
Questions into experiments.
Machine Learning Security
ADS-C: Antidistillation Sampling for Classification
An inference-time defense against model extraction that preserves every served top-1 prediction while degrading the supervision available to a distilling student.
−29.6 pp
distilled-student accuracy on CIFAR-10
0 pp served top-1 accuracy cost in the reported experiment
Multi-Agent Reinforcement Learning
Learning to Recharge a UAV Network
A custom environment for energy-constrained aerial recharging—and an investigation of the reward-induced “lazy agent” behavior that made inactivity more attractive than mission-critical work.
03 / PUBLICATIONS
Selected publications
How Do Large Language Models Judge Social Attraction? Evidence from Theory-Grounded Persona Ratings Across Multiple LLMs and Humans
arXiv · Preprint
BibTeX for How Do Large Language Models Judge Social Attraction? Evidence from Theory-Grounded Persona Ratings Across Multiple LLMs and Humans
@misc{mahmud2026socialattraction,
title = {How Do Large Language Models Judge Social Attraction? Evidence from Theory-Grounded Persona Ratings Across Multiple LLMs and Humans},
author = {Hasan Mahmud and Khawaja Abaid Ullah and Mohammad Javad Khojasteh and Jamison Heard and Prabu David},
year = {2026},
eprint = {2608.09717},
archivePrefix = {arXiv},
primaryClass = {cs.CL},
url = {https://arxiv.org/abs/2608.09717}
}
ADS-C: Antidistillation Sampling for Classification
arXiv · Preprint
BibTeX for ADS-C: Antidistillation Sampling for Classification
@misc{ullah2026adsc,
title = {ADS-C: Antidistillation Sampling for Classification},
author = {Khawaja Abaid Ullah and Mohammad Javad Khojasteh},
year = {2026},
eprint = {2607.15467},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2607.15467}
}
04 / ENGINEERING
I also build things.
Open-source tools and ML infrastructure, informed by work in research and production environments.
Teras
A multi-backend library for deep learning on tabular data, built on Keras 3. Architectures for prediction, imputation, and synthetic-data generation across JAX, PyTorch, and TensorFlow.
Keras · JAX · PyTorch · TensorFlow
Keras
Open-source contributions to multi-backend ML infrastructure, including Beta and Binomial random-sampling APIs and backend consistency fixes.
Open source · Multi-backend APIs
05 / ABOUT
Research, with an engineering instinct.
I am a PhD student in Electrical and Computer Engineering at Rochester Institute of Technology and a Graduate Research Assistant in the Khojasteh Autonomous Systems Laboratory, supervised by Dr. Mohammad Javad Khojasteh. My research spans machine learning security and autonomous systems: anti-distillation defenses against model extraction, and coordinated multi-agent robots for search-and-rescue in disaster-struck regions.
I received my M.S. in Artificial Intelligence from RIT in 2026 as a Fulbright Scholar. During my master's, I spent summer 2025 in Palo Alto as a Software Development Engineer Intern for AI/ML at Amazon Web Services. I previously studied computer science at the University of Narowal in Pakistan. Alongside my research, I build and contribute to open-source ML tools, including Teras and Keras.
06 / RECENT
Along the way
Released ADS-C, an antidistillation defense for classification.
Began PhD research in Electrical & Computer Engineering at RIT.
Completed my M.S. in Artificial Intelligence at RIT as a Fulbright Scholar.
Completed an AI/ML software engineering internship at AWS.