Search NASA⌕ Search

SEARCH · Search NASA

Results for “paradigms”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 397 records · Page 22
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Spectral kernel machines with electrically tunable photodetectors

Spectral machine vision collects spectral and spatial information as three-dimensional hypercubes and digitally processes them, which causes a data bottleneck, limiting power efficiency, frame rate, and spectral-spatial resolution. This work introduces spectral kernel machines (SKMs) to overcome these bottlenecks. SKM directly compresses spectral analysis through the output photocurrent and learns from example objects to identify and classify new samples in a "sniff-and-seek" mode. We experimentally demonstrated SKMs with electrically tunable bipolar black phosphorus-molybdenum disulfide (bP-MoS2) photodiodes in the near- and mid-infrared band and silicon photoconductors in the visible band, performing versatile intelligent tasks from chemometrics to semiconductor metrology. This architecture consumed substantially less power and was more than an order of magnitude faster than existing solutions for hyperspectral image analysis, defining an intelligent imaging and sensing paradigm with intriguing possibilities.

Zhang, Dehui↗
DOE OSTI2026

Knowledge gaps for neuromorphic ionic computing

BACKGROUND Neuromorphic computing, inspired by the human brain’s ability to process information efficiently, represents a transformative approach to computation. In this Review, we explore the emerging field of neuromorphic ionic computing, which leverages ionic conduction and coupling to mimic neural processes, and identify critical knowledge gaps that must be addressed to realize its full potential. A central theme of the discussion is energy efficiency, a challenge that is both a limitation and an opportunity for this technology. Although complementary metal-oxide semiconductor (CMOS)–based neuromorphic technologies have made strides in scaling to billions of neurons and are increasingly applied in artificial intelligence and numerical computing, they remain orders of magnitude behind the human brain in terms of connectivity and energy efficiency. Neuromorphic ionic computing promises to overcome these limitations by leveraging the distinct architectural and operational principles of the brain. Our brains achieve this energy efficiency by combining several key features: using the same network elements to store and process information; using an incredibly complex and massively interconnected three-dimensional (3D) network of locally active elements that enables sparsity, robustness in the presence of noise, adaptation, and life-long learning; computing at comparatively low voltage and frequency; and last, taking advantage of a plethora of ions and small molecules as information carriers. Here, we propose that ionic computing systems can take advantage of similar features to achieve substantial gains in energy efficiency. ADVANCES Since the first reports of neuromorphic ionic behavior in nanofluidic channels, we have witnessed an explosion of reports that used ionic devices to produce synaptomimetic behaviors. However, achieving the goals of ionic computing requires not only implementation of much more sophisticated device functionality but also overcoming fundamental barriers in materials science, device architecture, and system integration. Current ionic devices, even those incorporating state-of-the-art materials, still suffer from limited functionality and stability, which restrict their performance and increase energy demands. Developing new materials with enhanced ionic properties is essential to overcome these limitations. Similarly, the design of neuromorphic devices must evolve to leverage the particular advantages of ionic processes. Existing architectures often follow a single-information-carrier logic of conventional electronics or are constructed of mesoscale fluidics, failing to capitalize on the energy-efficient mechanisms inherent to ionic systems or implement the multiple-information-carrier paradigm. Current neuromorphic chips focus on large-scale networks of analog memory elements based on mechanisms such as charge trap (flash), filamentary, phase change, or spin, which are built on top of a network of artificial CMOS neurons. Although such prototype networks have achieved impressive performance, it is difficult to envision how they can implement the key features such as massive connectivity, sophisticated plasticity, adaptability, sparsity, and “multichromatic” computing. Although small-scale devices have demonstrated promising results, integrating them, maintaining energy efficiency, and implementing temperature control as systems grow in complexity and size to computationally relevant scale remain major hurdles. Furthermore, interfacing neuromorphic ionic devices with existing computing technologies presents technical and conceptual challenges that will require innovative approaches that combine insights from neuroscience, materials science, and engineering. OUTLOOK Despite these challenges, the potential impact of neuromorphic ionic computing is profound with potential applications ranging from artificial intelligence to robotics and beyond. We also argue that neuromorphic ionic computing systems should not, at least in the beginning, compete with CMOS technologies but rather should focus on applications that require extreme energy efficiency with chemical and/or biological compatibility, such as biomedical applications (for example, brain-computer interfaces), environmental monitoring, and agricultural and food applications. Ultimately, this Review highlights the crucial role of interdisciplinary collaboration in advancing the field. Neuromorphic ionic computing is not merely a technological innovation; it represents a substantial step toward sustainable computation, aligning with the growing demand for energy-conscious solutions in a world that is increasingly reliant on data and computation.

Neuromorphic↗
DOE OSTI2025

Physiological roles of an Acinetobacter -specific σ factor

ABSTRACT The Gram-negative pathogen Acinetobacter baumannii is considered an “urgent threat” to human health due to its propensity to become antibiotic resistant. Understanding the distinct regulatory paradigms used by A. baumannii to mitigate cellular stresses may uncover new therapeutic targets. Many γ-proteobacteria use the extracytoplasmic function (ECF) σ factor, RpoE, to invoke envelope homeostasis networks in response to stress. Acinetobacter species contain the poorly characterized ECF “SigAb”; however, it is unclear if SigAb has the same physiological role as RpoE. Here, we show that SigAb is a metal stress-responsive ECF that appears unique to Acinetobacter species and distinct from RpoE-like ECFs. We combine promoter mutagenesis, motif scanning, and chromatin immunoprecipitation-sequencing (ChIP-seq) to define the direct SigAb regulon, which consists of genes encoding SigAb itself, the stringent response mediator, RelA, and the uncharacterized small RNA, “SabS.” However, RNA-seq of strains overexpressing SigAb revealed a large, indirect regulon containing hundreds of genes. Metal resistance genes are key elements of the indirect regulon, as CRISPRi knockdown of sigAb or sabS resulted in increased copper sensitivity and excess copper-induced SigAb-dependent transcription. Furthermore, we found that two uncharacterized genes in the sigAb operon, “ aabA ” and “ aabB ,” have anti-SigAb activity. Finally, employing a targeted Tn-seq approach that uses CRISPR-associated transposons, we show that sigAb , aabA , and aabB are important for fitness even during optimal growth conditions. Our work reveals new physiological roles for SigAb and SabS, provides a novel approach for assessing gene fitness, and highlights the distinct regulatory architecture of A. baumannii . IMPORTANCE Acinetobacter baumannii is a hospital-acquired pathogen, and many strains are resistant to multiple antibiotics. Understanding how A. baumannii senses and responds to stress may uncover novel routes to treat infections. Here, we examine how the Acinetobacter -specific transcription factor, SigAb, mitigates stress. We find that SigAb directly regulates only a small number of genes, but indirectly controls hundreds of genes that have substantial impacts on cell physiology. We show that SigAb is required for maximal growth, even during optimal conditions, and is acutely required during growth in the presence of elevated copper. Given that copper toxicity plays roles in pathogenesis and on copper-containing surfaces in hospitals, we speculate that SigAb function may be important in clinically relevant contexts.

Bacon, Emily E. (ORCID:0000000180907689)↗
DOE OSTI2026

High-Order Mesh r-Adaptivity with Tangential Relaxation and Guaranteed Mesh Validity

High-order meshes are crucial for achieving optimal convergence rates in curvilinear domains, preserving symmetry, and aligning with key flow features in moving mesh simulations [1], but their quality is challenging to control. In prior work, we have developed techniques based on Target-Matrix Optimization Paradigm (TMOP) to adapt a given high-order mesh to the geometry and solution of the partial differential equation (PDE) [2, 3]. Here, we extend this framework to address two key gaps in the literature for highorder mesh 𝑟-adaptivity. First, we introduce tangential relaxation on curved surfaces using solely the discrete mesh representation, eliminating the need for access to underlying geometry (e.g., CAD model). Second, we ensure a continuously positive Jacobian determinant throughout the domain. This determinant positivity is essential for using the high-order mesh resulting from 𝑟-adaptivity with arbitrary quadrature schemes in simulations. The proposed approach is demonstrated to be robust using a variety of numerical experiments.

Mathematics and Computing↗
DOE OSTI2024

On the Training and Generalization of Deep Operator Networks

Here, we present a novel training method for deep operator networks (DeepONets), one of the most popular neural network models for operators. DeepONets are constructed by two subnetworks, namely the branch and trunk networks. Typically, the two subnetworks are trained simultaneously, which amounts to solving a complex optimization problem in a high dimensional space. In addition, the nonconvex and nonlinear nature makes training very challenging. To tackle such a challenge, we propose a two-step training method that trains the trunk network first and then sequentially trains the branch network. The core mechanism is motivated by the divide-and-conquer paradigm and is the decomposition of the entire complex training task into two subtasks with reduced complexity. Therein the Gram–Schmidt orthonormalization process is introduced which significantly improves stability and generalization ability. On the theoretical side, we establish a generalization error estimate in terms of the number of training data, the width of DeepONets, and the number of input and output sensors. Numerical examples are presented to demonstrate the effectiveness of the two-step training method, including Darcy flow in heterogeneous porous media.

deep operator networks↗
DOE OSTI2025

Bias-Variance Trade-Off in Physics-Informed Neural Networks with Randomized Smoothing for High-Dimensional PDEs

Physics-Informed Neural Networks (PINNs) have triggered a paradigm shift in scientific computing, leveraging mesh-free properties and robust approximation capabilities. While proving effective for low-dimensional partial differential equations (PDEs), the computational cost of PINNs remains a hurdle in high-dimensional scenarios. This is particularly pronounced when computing high-order and high-dimensional derivatives in the physics-informed loss. Randomized Smoothing PINN (RS-PINN) introduces Gaussian noise for stochastic smoothing of the original neural net model, enabling the use of Monte Carlo methods for derivative approximation, which eliminates the need for costly automatic differentiation. Despite its computational efficiency, especially in the approximation of high-dimensional derivatives, RS-PINN introduces biases in both loss and gradients, negatively impacting convergence, especially when coupled with stochastic gradient descent (SGD) algorithms. We present a comprehensive analysis of biases in RS-PINN, attributing them to the nonlinearity of the Mean Squared Error (MSE) loss as well as the intrinsic nonlinearity of the PDE itself. We propose tailored bias correction techniques, delineating their application based on the order of PDE nonlinearity. The derivation of an unbiased RS-PINN allows for a detailed examination of its advantages and disadvantages compared to the biased version. Specifically, the biased version has a lower variance and runs faster than the unbiased version, but it is less accurate due to the bias. To optimize the bias-variance trade-off, we combine the two approaches in a hybrid method that balances the rapid convergence of the biased version with the high accuracy of the unbiased version. In addition to methodological contributions, we present an enhanced implementation of RS-PINN. Extensive experiments on diverse high-dimensional PDEs, including Fokker-Planck, Hamilton-Jacobi-Bellman (HJB), viscous Burgers’, Allen-Cahn, and Sine-Gordon equations, illustrate the bias-variance trade-off and highlight the effectiveness of the hybrid RS-PINN. Empirical guidelines are provided for selecting biased, unbiased, or hybrid versions, depending on the dimensionality and nonlinearity of the specific PDE problem.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Robust Optimal Experimental Design of Infinite-Dimensional Bayesian Nonlinear Inverse Problems

Abstract. We consider robust optimal experimental design (ROED) for nonlinear Bayesian inverse problems governed by partial differential equations (PDEs). An optimal design is one that maximizes some utility quantifying the quality of the solution of an inverse problem. However, the optimal design is dependent on elements of the inverse problem such as the simulation model, the prior, or the measurement error model. ROED aims to produce an optimal design that is aware of the additional uncertainties encoded in the inverse problem and remains optimal even after variations in them. We follow a worst-case scenario approach to develop a new framework for robust optimal design of nonlinear Bayesian inverse problems. The proposed framework (a) is scalable and designed for infinite-dimensional Bayesian nonlinear inverse problems constrained by PDEs; (b) develops efficient approximations of the utility, namely the expected information gain; (c) employs eigenvalue sensitivity techniques to develop analytical forms and efficient evaluation methods of the gradient of the utility with respect to the uncertainties against which we wish to be robust; and (d) employs a probabilistic optimization paradigm that properly defines and efficiently solves the resulting combinatorial max-min optimization problem. The effectiveness of the proposed approach is illustrated for optimal sensor placement problem in an inverse problem governed by an elliptic PDE.

Chowdhary, Abhijit↗
DOE OSTI2026

A roadmap toward scaling, reasoning and self-evolving foundation models for nuclear and particle physics

Foundation models have revolutionized artificial intelligence, with Large Language Models demonstrating unprecedented capabilities in multimodal understanding, reasoning and tool use. Nuclear and particle physics stands at a critical juncture where similar transformative potential awaits realization. The field generates exabytes of experimental data, exascale simulations, and decades of theoretical insights — yet these remain largely disconnected from modern Artifical Intelligence (AI) capabilities, with most physics AI applications confined to narrow, task-specific models that suffer from domain shifting when applied to real experimental data. We present a roadmap for FM4NPP (Foundation Model for Nuclear and Particle Physics), systematically scaling from current proof-of-concept models to trillion-parameter architectures capable of autonomous discovery. Our approach advances three critical frontiers: unified data infrastructure integrating detector data, scientific knowledge and computational tools across global facilities; multi-facility foundation models enabling cross-experiment knowledge transfer and accelerated discovery; and agentic AI capabilities for reasoning and autonomous tool use. The resulting self-evolving FM4NPP will transform physics research by converting time-intensive data analysis, theory derivation and computational bottlenecks into rapid AI–human collaborative discovery. This paradigm shift promises to fundamentally accelerate scientific progress in nuclear and particle physics, enabling researchers to focus on high-level insights while AI handles routine analysis and explores vast parameter spaces beyond human capacity.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Jet substructure

Jet substructure observables hold the keys to identifying the inner working of the quark–gluon plasma through its imprints in jet modification patterns. Because of the multi-scale nature of jet evolution, the strategy has focused on developing jet reclustering-based and jet shape-based observables to specifically probe different stages of jet evolution, which provide multi-dimensional quantification of jet particle distribution phase spaces. The developments of jet substructure in the past few years allowed us to explore the correlations among jet observables. The richer information beyond single-observable modifications could give provides a crucial step beyond the energy loss paradigm. This review surveys the theory and phenomenology of modern jet substructure observables, which help constraining the energy scales of medium interactions and identifying heavy flavor quarks within jets. Also, the peculiar soft activities point to more refined studies of medium responses to jets.

Physics↗
DOE OSTI2024

AGS-GNN: Attribute-guided Sampling for Graph Neural Networks

We propose AGS-GNN, a novel attribute-guided sampling algorithm for Graph Neural Networks (GNNs) that exploits node features and connectivity structure of a graph while simultaneously adapting for both homophily and heterophily in graphs. (In homophilic graphs vertices of the same class are more likely to be connected, and vertices of different classes tend to be linked in heterophilic graphs.) While GNNs have been successfully applied to homophilic graphs, their application to heterophilic graphs remains challenging. The best-performing GNNs for heterophilic graphs do not fit the sampling paradigm, suffer high computational costs, and are not inductive. We employ samplers based on feature-similarity and feature-diversity to select subsets of neighbors for a node, and adaptively capture information from homophilic and heterophilic neighborhoods using dual channels. Currently, AGS-GNN is the only algorithm that we know of that explicitly controls homophily in the sampled subgraph through similar and diverse neighborhood samples. For diverse neighborhood sampling, we employ submodularity, which was not used in this context prior to our work. The sampling distribution is pre-computed and highly parallel, achieving the desired scalability. Using an extensive dataset consisting of 35 small (<=100K nodes) and large (>100K nodes) homophilic and heterophilic graphs, we demonstrate the superiority of AGS-GNN compare to the current approaches in the literature. AGS-GNN achieves comparable test accuracy to the best-performing heterophilic GNNs, even outperforming methods using the entire graph for node classification. AGS-GNN also converges faster compared to methods that sample neighborhoods randomly, and can be incorporated into existing GNN models that employ node or graph sampling.

artificial intelligence↗
DOE OSTI2024

Emerging Technologies for Privacy Preservation in Energy Systems

This study explores the intersection of digitalization and privacy within the energy sector, focusing on the emerging challenges and opportunities presented by integrating Distributed Energy Resources (DERs) and advanced metering infrastructure. The need for robust digital privacy measures has become crucial as the energy industry evolves towards a more decentralized, digitalized, and decarbonized future. This study delves into four cutting-edge privacy-preserving technologies—Homomorphic Encryption (HE), Secure Multiparty Computation (SMPC), Differential Privacy (DP), and Federated Learning (FL)—each offering unique solutions to safeguard consumer data by increasing digital connectivity and data exchange. Through a detailed examination of these methods, the study explains how each technology operates, its applications within the energy sector, and the specific privacy challenges it addresses. Homomorphic Encryption allows for secure computations on encrypted data, enabling data analysis without compromising privacy. Secure Multiparty Computation enables collaborative data analysis across different entities while protecting the confidentiality of the inputs. Differential Privacy introduces randomness into the assembled data set, preventing the identification of individual records in statistical databases. Lastly, Federated Learning offers a paradigm shift in data analysis, where machine learning models are trained at the edge, minimizing the centralization of sensitive data. The research underscores the significance of implementing these privacy-enhancing technologies to comply with strict data protection regulations, foster consumer trust, and enhance the security of the energy infrastructure. By providing a comprehensive overview of these methodologies and their practical implications for the energy sector, this study aims to contribute to the ongoing discourse on digital privacy, offering insights into how the energy industry can navigate the complexities of data privacy in the digital age.

Cali, Umit↗
DOE OSTI2024

Virtual Time III, Part 3: Throttling and Message Cancellation

This is Part 3 of a trio of papers that unify in a natural way the two historically distinct parallel discrete event synchronization paradigms, optimistic and conservative, combining the best properties of both into a single framework called Unified Virtual Time (UVT). In this part, we survey the synchronization effects that can be achieved by restricting to corner cases the relationships permitted among the control variables, GVT, CVT, TVT, and LVT, which were defined in Part 1. Here we also survey various throttling policies from the literature and describe how they can be implemented in UVT by controlling the value of TVT, including policies that can take advantage of rollback in addition to LP blocking. A significant result is a new category of efficient and higher precision throttling algorithms for optimistic execution that are based on optimistic lookahead, defined in a way that is symmetric to what we now call the conservative lookahead information that is traditionally used for conservative synchronization. Finally, we present a novel algorithm allowing the choice between lazy and aggressive cancellation to be made on a message-by-message basis using either external logic expressed in the model code, or policy code internal to the simulator, or a mixture of both.

throttling↗
DOE OSTI2025

ASDF: A Compiler for Qwerty, a Basis-Oriented Quantum Programming Language

Qwerty is a high-level quantum programming language built on bases and functions rather than circuits. This new paradigm introduces new challenges in compilation, namely synthesizing circuits from basis translations and automatically specializing adjoint or predicated forms of functions. This paper presents ASDF, an open-source compiler for Qwerty that answers these challenges in compiling basis-oriented languages. Enabled with a novel high-level quantum IR implemented in the MLIR framework, our compiler produces OpenQASM 3 or QIR for either simulation or execution on hardware. Our compiler is evaluated by comparing the fault-tolerant resource requirements of generated circuits with other compilers, finding that ASDF produces circuits with comparable cost to prior circuit-oriented compilers.

Adams, Austin J [Georgia Tech]↗
DOE OSTI2025

Quantum Transfer Learning to Boost Dementia Detection

Dementia is a devastating condition with profound implications for individuals, families, and healthcare systems. Early and accurate detection of dementia is critical for timely intervention and improved patient outcomes. While classical machine learning and deep learning approaches have been explored extensively for dementia prediction, these solutions often struggle with high-dimensional biomedical data and large-scale datasets, quickly reaching computational and performance limitations. To address this challenge, quantum machine learning (QML) has emerged as a promising paradigm, offering faster training and advanced pattern recognition capabilities. This work aims to demonstrate the potential of quantum transfer learning (QTL) to enhance the performance of a weak classical deep learning model applied to a binary classification task for dementia detection. Besides, we show the effect of noise on the QTL-based approach, investigating the reliability and robustness of this method. Using the OASIS 2 dataset, we show how quantum techniques can transform a suboptimal classical model into a more effective solution for biomedical image classification, highlighting their potential impact on advancing healthcare technology.

Bhowmik, Sounak [University of Tennessee, Knoxvill↗
DOE OSTI2025

eCounter: Inline Per-IP Network Monitoring at Millisecond Resolution via eBPF

Scientific data acquisition (SciDAQ) systems are shifting from archive-based workflows to streaming paradigms, where real-time, fine-grained network monitoring becomes essential. While P4-enabled devices offer per-packet in-band observability, they require specialized switches and routers. Host-side tools like Prometheus exporters lack sufficient temporal granularity. To bridge this gap, we present eCounter, a lightweight, hardware-agnostic, inline telemetry agent built on extended Berkeley Packet Filter (eBPF). eCounter captures per-interface ingress and egress traffic, categorized by IP address and protocol, at millisecond to sub-millisecond resolution. In a 100 Gbps environment, it continuously exports up to 3,257 time-series bins per second with only 4% CPU utilization at a 35¿KiB/s data rate. We evaluate eCounter across diverse NIC MTU settings, hook types, CPU architectures and operating systems, and observed negligible impact on concurrent high-throughput streaming applications. Complexity analysis confirms that it can be readily scaled to distributed SciDAQ deployments.

Mei, Xinxin [Computational Sciences and Technology↗
DOE OSTI2025

Data Assimilation for Robust UQ Within Agent-Based Simulation on HPC Systems

Agent-based simulation provides a powerful tool for in silico system modeling. However, these simulations do not provide built-in methods for uncertainty quantification (UQ). Within these types of models a typical approach to UQ is to run multiple realizations of the model then compute aggregate statistics. This approach is limited due to the compute time required for a solution. When faced with an emerging biothreat, public health decisions need to be made quickly and solutions for integrating near real-time data with analytic tools are needed. We propose an integrated Bayesian UQ framework for agent-based models based on sequential Monte Carlo sampling. Given streaming or static data about the evolution of an emerging pathogen this Bayesian framework provides a distribution over the parameters governing the spread of a disease through a population. These estimates of the spread of a disease may be provided to public health agencies seeking to abate the spread. By coupling agent-based simulations with Bayesian modeling in a data assimilation, our proposed framework provides a powerful tool for modeling dynamical systems in silico. We propose a method which reduces model error and provides a range of realistic possible outcomes. Moreover, our method addresses two primary limitations of ABMs: the lack of UQ and an inability to assimilate data. Our proposed framework combines the flexibility of an agent-based model with UQ provided by the Bayesian paradigm in a workflow which scales well to HPC systems. We provide algorithmic details and results on a simulated outbreak with both static and streaming data.

Spannaus, Adam [ORNL] (ORCID:0000000225213657)↗
DOE OSTI2025

BuildingQA: A Benchmark for Natural Language Question Answering over Building Knowledge Graphs

Graph-based representations of building metadata using ontologies like Brick are vital for smart building applications, but querying them remains a challenge for practitioners. Knowledge Graph Question Answering (KGQA) systems, meant to retrieve answers from natural language questions, traditionally require large-scale training data, making them ill-suited for the specialized and data-scarce building domain. The advent of Large Language Models (LLMs) offers a paradigm shift, enabling zero-shot natural language querying without building/domain-specific training. Yet, there is no standardized benchmark for building-specific KGQA which can guide and validate research in this area. To address this gap, our work makes three primary contributions. First, we introduce the BuildingQA Benchmark Dataset, constructed through a multi-stage process of collecting practitioner data, augmenting it with LLMs for linguistic diversity, and curating a final set of 188 questions across 4 buildings. Second, we characterize the benchmark's complexity and ambiguity, introducing a novel method to quantify its "lexical gap" and providing a four-stage diagnostic framework for analyzing how systems fail. Third, we benchmark zero-shot LLM-powered KGQA systems to establish baseline performance and analyze their failure modes. Our evaluation reveals that top-performing systems achieve a maximum F1 score of only 0.38. This result does not indicate a failure of these powerful systems, but rather underscores the unique challenges posed by our benchmark. It demonstrates a critical performance gap, showing that current methods successful on general KGs struggle with the specific lexical and structural nuances of the building domain. BuildingQA1 thus provides the benchmark dataset and foundational analysis needed to drive the development of novel, domain-aware methods required to unlock the use of semantic data in buildings.

Mulayim, Ozan Baris↗
DOE OSTI2025

Position-Enhanced Gradient Attack (PEGA) on Medical Language Models

Federated Learning (FL) enables collaborative training of language models on sensitive clinical notes without sharing the data. However, this paradigm is vulnerable to gradient inversion attacks that can reconstruct private data from shared gradients. We find that state-of-the-art attacks are less effective in the medical domain, failing to overcome the unique challenges posed by its specialized vocabulary and unstructured format. To address this, we introduce the Position-Enhanced Gradient Attack (PEGA), a novel attack that makes gradients position-aware by optimizing token and position embeddings simultaneously. PEGA employs two key innovations: a periodic sorting of positional embeddings to resolve token order ambiguity and a late-stage embedding replacement strategy to correct hard-to-recover critical tokens. To evaluate the leakage of sensitive data more directly, we also propose the Unified PHI-Recall (UPHI), a new metric measuring the recovery of Protected Health Information. Experiments on the MIMIC-III dataset show that PEGA significantly outperforms leading attacks like TAG and LAMP, particularly in its ability to reconstruct identifiable patient information, exposing a more severe and nuanced privacy risk in federated medical NLP.

Xu, Nuo [University of Minnesota]↗