DISys Lab
Our research is proudly supported by
Copyright DISys Lab 2024
11/17/25
Congratulations to Lucius for the acceptance of his paper on distributed Primal–Dual Hybrid Gradient (PDHG) methods to the SIAM Conference on Parallel Processing. His work develops scalable, parallelizable PDHG formulations designed for emerging in-memory and heterogeneous computing architectures, enabling significant reductions in latency and communication overhead for large-scale optimization problems.10/27/25
Paritosh presented the DISys Lab’s latest work on zero-knowledge cyberattack detection for regulatory compliance in critical infrastructure systems at the IISE Annual Conference. The talk introduced a scalable detection framework that allows regulators to verify alarm correctness without gaining access to sensitive operational data from utilities. Building on residual-based statistical hypothesis testing for state-space models, the framework incorporates a two-pronged zero-knowledge architecture enforcing both temporal coherence of system dynamics and statistical consistency of detection tests. This design enables regulators to validate cyberattack alarms in a privacy-preserving manner, addressing one of the key barriers to modernizing compliance standards for industrial control systems. The approach demonstrates strong theoretical soundness and zero-knowledge guarantees, and its feasibility was validated using real-world datasets from critical infrastructure networks.You can read more about it here
07/22/25
A new interdisciplinary research effort at Oklahoma State University is exploring how biofabrication facilities can collaboratively design personalized biomedical constructs while preserving data privacy. The work, supported by the National Science Foundation, focuses on enabling the secure exchange of design insights and manufacturing strategies between facilities—without revealing sensitive patient information or proprietary fabrication data. The project addresses a central challenge in personalized medicine: 3D-printing patient-specific tissues often requires extensive trial-and-error, but sharing such manufacturing knowledge between sites raises privacy and confidentiality concerns. The OSU team has developed a decentralized learning framework that incorporates differential-privacy protections to enable meaningful collaboration across institutions while maintaining strict privacy guarantees. The research team includes Professors Akash Deep, Srikanthan Ramesh, and Paritosh Ramanan, who together aim to advance secure biofabrication pipelines for next-generation personalized health care solutions. Their work highlights OSU’s growing leadership in digital manufacturing, biodesign, and privacy-preserving computational methods.You can read more about it here
06/01/25
DISys Lab is pleased to announce that the SplitVAEs paper has received the 2nd Runners Up in the Best Student Paper Award for its innovative contributions to decentralized scenario generation. SplitVAEs enables stakeholders to collaboratively train generative models without sharing raw data, addressing key challenges in data residency, privacy, and distributed modeling. The framework demonstrates how decentralized VAEs can generate high-quality, distribution-aligned scenarios suitable for robust stochastic optimization across industrial and energy domains. Congratulations to Lucius for this well-deserved recognition of his leadership and contributions.You can read more about it here
05/20/25
DISys Lab members presented two research efforts centered on advancing decarbonization and data-driven planning for industrial and power system operations. The first talk, presented by Richard Reed, Saba Ghasemi, Paritosh Ramanan, and Zheyu Jiang, discussed decentralized operations planning for electrified chemical plants integrated with renewable-driven transmission systems. The team presented a decentralized mixed-integer optimization strategy using iterative ADMM along with a relaxation-to-integer transition to achieve consensus across stakeholders. Their results demonstrated that decentralized unit commitment and process heating operations can meet demand while preserving stakeholder privacy and maintaining performance comparable to centralized benchmarks. The second presentation, delivered by Huynh Quang Nguyen Vo, Richard Reed, Saba Ghasemi Naraghi, and Zheyu Jiang, explored decentralized importance sampling using variational autoencoders to generate high-fidelity industrial carbon emissions scenarios. Leveraging decentralized VAE-based methods, the team demonstrated how generated scenarios capture spatiotemporal interdependencies across multiple chemical plants while respecting privacy constraints inherent to siloed datasets. These scenarios, supported by importance weighting, were shown to improve stochastic optimization workflows for planning decarbonized industrial and power system operations.You can read more about it here
04/23/25
Dr. Ramanan will be attending ICLR 2025 in Singapore. This is joint work with Ayush Mohanty and Nagi Gebraeel at Georgia Tech and follows from our experiences building large scale decentralized machine learning models as part of the NASA HOME STRI projectYou can read more about it here
04/15/25
Ricky's research focuses on developing decentralized optimization methods for operations planning in electrified chemical process heating systems integrated with renewable-driven power grids. His work leverages iterative ADMM and relaxation-to-integer transitions to coordinate decision-making across multiple stakeholders while preserving operational privacy. Congratulations to Ricky on an excellent defense and impactful contributions to the DISys Lab's research on industrial decarbonization.04/07/25
The InCySe Hackathon 2025 has officially begun, running from April 7th to April 17th. Both the Identify–Build–Demonstrate (IBD) and Find the Flag (FTF) challenges are live, inviting students to showcase their skills in decentralized systems and industrial cybersecurity. The IBD Challenge encourages participants to design and build decentralized applications using technologies such as Ethereum and IPFS for industrial use cases including anomaly detection, energy trading, and decentralized auctions. Teams will present a poster and working prototype at the IEM Research Symposium on April 17th. The Find the Flag Challenge tasks teams with uncovering vulnerabilities in a hosted critical-infrastructure-inspired IT system. Participants gain hands-on experience with security audits, REST API concepts, authentication mechanisms, and cyber-hygiene practices for protecting industrial control systems. Registration opened on March 12th, with tutorials released on March 19th. The submission window opens April 7th, the FTF challenge releases April 9th, and all final submissions are due April 16th. Winners will be announced at the IEM Awards Banquet on April 18th. Prizes totaling $1400 are available across both competitions, with awards for winners and runners-up in the IBD and FTF categories. Tutorials and resources covering blockchain, cybersecurity fundamentals, and relevant Python tools are provided to support participants. For questions, participants may contact the organizing team at incyse-hackathon@okstate.edu.You can read more about it here
03/14/25
Dr. Ramanan presented a differential privacy framework designed to help regulators verify cyberattack detection across multi-utility energy networks without exposing sensitive operational data.You can read more about it here
10/19/24
Lucius and Ricky are going to be presenting their research pertaining to Importance Weighted Autoencoders (TA76: Computational Perspectives for Promoting Industrial Decarbonization) and decentralized formulations for decarbonized process heating. Mohaimanul is presenting his work on decentralized methods of scenario generation (WC20: Application of Federated and Distributed Learning Frameworks).You can read more about it here
09/30/24
The grant from the Secure and Trustworthy Cyberspace (SaTC) program of NSF explores privacy preserving, decentralized detection of attacks in multi-stakeholder infrastructure systems. Part of the research will look into statistical guarantees and cryptographically secure proofs of computation regarding detection outcomes.You can read more about it here
07/12/24
Our paper entitled "Catalyzing Deep Decarbonization : Federated Battery Diagnosis and Prognosis for Better Data Management in Energy Storage Systems" has been accepted in Cell Reports Physical Science. In this paper we explore an understated benefit of decentralized federated learning paradigms that enables small and medium sized enterprises (SMEs) to reduce the barrier of entry for accessing sophisticated machine learning insights. We demonstrate SMEs in the energy storage industry can obtain high quality, prognostics outcomes in a fully decentralized fashion without the need to move their data. This is joint work with Dr. Murat Yildirim from Wayne State University and Drs. Susan Babinec, Feng Qiu and Noah Paulson from Argonne National Labs.07/01/24
The NSF EAGER grant that focuses on developing privacy focused decentralized methodologies to securely and safely integrate chemical process heating systems with existing renewable-driven, electric transmission systems. This is joint work with Dr. Zheyu Jiang from School of Chemical Engineering at OSU.You can read more about it here
06/10/24
Our paper "The Lynchpin of In-Memory Computing: A Benchmarking Framework for Vector-Matrix Multiplication in RRAMs" has been accepted in ACM ICONS 2024. In this paper, we develop a novel, scalable framework for simulating matrix vector multiplications in analog space on neuromorphic computational systems. Our framework can be used to develop higher order optimization and AI libraries that can be executed and simulated on neuromorphic accelerators. This is joint work with Dr. Murat Yildirim, Gozde Tutuoncuoglu from Wayne State University.05/18/24
Dr. Ramanan's session entitled "Privacy and Security in Power Systems" included talks on decentralized planning and operations in the power system domains while also including latest DISys Lab research on decentralized scenario generation.12/05/23
Timman's research is aimed at removing computational and scalability barriers corresponding to the use of Variational Autoencoders (VAEs) for scenario generation for stochastic optimization problems.DISys Lab
Our research is proudly supported by
Copyright DISys Lab 2024