Anshuman Suri

San Francisco, CA

I recently joined as a Member of Technical Staff at DatologyAI! Previously, I was a postdoctoral fellow at Northeastern University, supervised by Alina Oprea. I obtained my PhD from University of Virginia, where I was extremely fortunate to be advised by Professor David Evans.

During my time in academia, I studied the privacy and security challenges of machine learning models, both as targets of attacks and as amplifiers of risk; exploring how models can be exploited and how they transform the security landscape of systems that use them.

news

Jul 14, 2026	Presented our work Toward a Principled Framework for Agent Safety Measurement at the FAgEn workshop at ICML in Seoul!
Dec 01, 2025	Awarded OpenPhilantrophy (now Coefficint Giving) Research Grant (along with co-PIs Alina and Cristina) to work on belief-falsification based poisoning for LLMs. Thank you, Coefficint Giving!
Jul 25, 2025	Awarded Lambda Research Grant to work on training-time poisoning defenses for LLMs. Thank you, Lambda!
Jan 24, 2025	Our blogpost talking about critical flaws in the evaluation of a recent EMNLP best-paper has been accepted to the ICLR blogpost track!
Nov 14, 2024	UVA Engineering covered a story on our work on evaluating Membership Inference Attacks on Large Language Models?.

latest posts

Jul 17, 2025	Advice for working on ML projects
Nov 26, 2024	Reassessing EMNLP 2024’s Best Paper: Does Divergence-Based Calibration for Membership Inference Attacks Hold Up?
Nov 12, 2024	My submission to the ETI Challenge

selected publications

PoisonForge: Task-Level Targeted Poisoning Benchmark for Instruction-Tuned LLMs

Luze Sun, Anshuman Suri, Harsh Chaudhari, Cristina Nita-Rotaru, and 1 more author

arXiv preprint arXiv:2605.23168, 2026

Bib PDF

@article{sun2026poisonforge,
  title = {PoisonForge: Task-Level Targeted Poisoning Benchmark for Instruction-Tuned LLMs},
  author = {Sun, Luze and Suri, Anshuman and Chaudhari, Harsh and Nita-Rotaru, Cristina and Oprea, Alina},
  journal = {arXiv preprint arXiv:2605.23168},
  year = {2026},
}

Toward Principled LLM Safety Testing: Solving the Jailbreak Oracle Problem

Shuyi Lin, Anshuman Suri, Alina Oprea, and Cheng Tan

In Proceedings of Machine Learning and Systems (MLSys), 2026

Bib PDF Code

@inproceedings{lin2026toward,
  title = {Toward Principled LLM Safety Testing: Solving the Jailbreak Oracle Problem},
  author = {Lin, Shuyi and Suri, Anshuman and Oprea, Alina and Tan, Cheng},
  booktitle = {Proceedings of Machine Learning and Systems (MLSys)},
  year = {2026}
}

Exploiting Leaderboards for Large-Scale Distribution of Malicious Models

Anshuman Suri^*, Harsh Chaudhari^*, Yuefeng Peng^*, Ali Naseh^*, and 2 more authors

In IEEE Symposium on Security and Privacy (S&P), 2026

Bib PDF Code

@inproceedings{suri2026exploiting,
  title = {Exploiting Leaderboards for Large-Scale Distribution of Malicious Models},
  author = {Suri, Anshuman and Chaudhari, Harsh and Peng, Yuefeng and Naseh, Ali and Oprea, Alina and Houmansadr, Amir},
  booktitle = {IEEE Symposium on Security and Privacy (S&P)},
  year = {2026}
}