RESEARCH / OVERVIEW
I work on anti-distillation defenses that protect deployed models against extraction attacks. My autonomous systems research aims to develop coordinated multi-agent robots for search-and-rescue in disaster-struck regions.
01 / Secure learning
Prediction APIs expose more than a label. I study how their probability distributions enable distillation-based replication, and how inference-time defenses can reduce that signal without changing the served top-1 prediction.
Model extraction · Antidistillation
02 / Autonomous multi-agent systems
My work in autonomous systems is driven by the goal of developing highly coordinated multi-agent robots for search-and-rescue in disaster-struck regions. I study reinforcement learning and coordination under energy, information, and communication constraints in support of that goal.
Reinforcement learning · Coordination
03 / Machine reasoning & optimization
Current directionI am exploring LLM-assisted combinatorial decision-making, especially where a problem is distributed across agents. This is an emerging direction. My public collaborative work on LLM evaluation provides a complementary perspective on the reliability of model judgments.
LLMs · Agents · Optimization
Machine Learning Security
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
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.
arXiv · Preprint
@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}
}
arXiv · Preprint
@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}
}
Interested in research, collaboration, or applied ML work?
Feel free to reach out ↗