2434 Osborn Dr · Dept. of Computer Science · Iowa State University · Ames, IA 50011 USA
Office: 232 Atanasoff Hall · Email: meisam@iastate.edu
Affiliation: Responsible Computing Lab, Dept. of Computer Science, Iowa State University (ISU)
© 2025 Meisam Mohammady
Dr. Meisam Mohammady is an Assistant Professor in the Department of Computer Science at Iowa State University (ISU), where his research focuses on developing responsible Machine Learning methods that are privacy-preserving, adversarially robust, and fair, leveraging tools such as Differential Privacy, Learning Theory, and Optimization, with applications in High Performance Computing (HPC), Federated Learning (FL), Networking, Anomaly Detection, and Private Retrieval. His research has been published in top-tier conferences and journals such as IEEE Symposium on Security and Privacy (IEEE S&P), ACM Conference on Computer and Communications Security (ACM CCS), ACM Transactions on Intelligent Systems and Technology (ACM TIST), IEEE Computer Security Foundations Symposium (IEEE CSF), IEEE Transactions on Dependable and Secure Computing (IEEE TDSC), ACM Transactions on Privacy and Security (ACM TOPS), and IEEE Transactions on Knowledge and Data Engineering (IEEE TKDE).
Prior to joining ISU, he was a Research Scientist at CSIRO’s Data61, Australia’s leading digital research network.
I am always looking for motivated students, visiting scholars/students, and undergraduate researchers.
Please email your application materials to Dr. Meisam Mohammady if you are interested in our research.
Graduate admission information can be found here.
🧭 Recent News
🧪 Groundbreaking Advances in Privacy-Preserving Machine Learning
We’re excited to share recent successes from our group in privacy-preserving deep learning, featured at top security venues. Our pioneering work on enabling the Laplace family in DP-SGD has delivered transformative results:
- PLRV-O: Advancing Differentially Private Deep Learning via Privacy-Loss Random Variable Optimization, accepted at ACM CCS 2025, introduces the first task-adaptive approach that jointly designs noise and clipping parameters for each deep learning task—defined by epochs, batch size, and privacy guarantee. It achieves up to 11% higher accuracy on ViT/CIFAR-10 and 40% on RoBERTa-Large/SST-2 compared to standard DP-SGD. 🏅 Congrats to Qin, Nicholas, Ayesha, and all co-authors!
- Lap₂: Revisiting Laplace DP-SGD for High Dimensions via Majorization Theory, accepted at IEEE CSF 2026, shows that Laplace noise can outperform Gaussian in DP-SGD by applying majorization theory, achieving 40–50% accuracy gains at the same privacy budget compared to standard Laplace approaches. 🏅 Congrats to Qin, Nicholas, Ayesha, and the team!
🧩 Private Federated Recommendation
FedSIG: Privacy-Preserving Federated Recommendation via Synthetic Interaction Generation, accepted at RAID 2025, introduces synthetic interaction generation for privacy-preserving recommendation.
🏅 Congrats to Thirasara and the team!
🚗 NSF Collaborative Grant Awarded
Collaborative grant Privacy-Preserving Collaborative Data Sharing for Intelligent Transportation awarded by NSF PDaSP (2025–2028) in partnership with UConn, UW, and IIT.
📊 High-Utility Data Sharing Framework
Towards Usability of Data with Privacy: A Unified Framework for Privacy-Preserving Data Sharing with High Utility, accepted at ASIACCS 2025.
🏅 Congrats to Chamikara and the team!
⚙️ Harmonized Differential Privacy in Federated Learning
Harmonizing Differential Privacy Mechanisms for Federated Learning: Boosting Accuracy and Convergence, accepted at CODASPY 2025.
🏅 Congrats to Shuya and the team!
🔐 User-Driven Local Differential Privacy
UD-LDP: A Technique for Optimally Catalyzing User-Driven Local Differential Privacy, to appear in FGCS 2025.
🏅 Congrats to Gnana and collaborators!
💡 Patent Granted
Utility-Optimized Differential Privacy System — U.S. Patent No. 12321478.
🏅 Congrats to Mengyuan and the team!
📘 Foundational Privacy Study
DPI: Ensuring Strict Differential Privacy for Infinite Data Streaming, appeared in IEEE S&P 2024.
🏅 Congrats to Shuya and all co-authors!