Search NASA⌕ Search

SEARCH · Search NASA

Results for “Deep ensemble”

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 253 records · Page 14
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Sentinel

Network intrusion detection systems (NIDS) are commonplace in network security but they frequently employ algorithms that are computational demanding requiring hardware and software with significant power requirements. Two examples of such resource-intensive algorithms used for network security are regular expression matching and broader signature pattern matching which are commonly used in deep packet inspection (DPI). Network security algorithms that have large power requirements may be a challenge for low-power internet-of-things (IoT) environments, which generally lack the power resources to implement complex security measures like computationally expensive DPI at the edge. Furthermore, IoT environments incorporating 5G standalone networks have network latency constraints beyond just power that make DPI at the edge even more difficult. Programmable logic is ideally suited for machine learning inference for DPI because of its deep instruction level parallelism and single-cycle memory access. Machine learning approaches for DPI have been explored before using the programmable logic of field programmable gate arrays (FPGA) as a potential solution for NIDS approaches that would be power-suitable for IoT. However, those previous programmable logic NIDS approaches utilize either a supervised or unsupervised learning model. Sentinel utilizes the ensemble of these two machine learning approaches known as a semi-supervised approach which has shown promise in NIDS implementations. Sentinel provides a programmable logic implementation of a semi-supervised approach for DPI which operates at much lower power and latency than a GPU implementation with negligible loss of accuracy due to quantization through a logistic regressor.

Anderson, MatthewW [Idaho National Laboratory (INL↗
DOE OSTI2026

Electronic and Structural Properties of the Polymer-Electrolyte Interphase in Electrochemically Doped Polymers

The advance of soft, polymer-based (photo)electrochemical energy transformation and storage applications requires a framework for polymer and electrolyte design that includes a deep understanding of the polymer–electrolyte interphase. Here, we report on the investigations of two napthalenediimide (NDI)–bithiophene (T2)-based semiconducting copolymers using computational modeling and in situ, ex situ, and operando techniques to reveal how changes in electrolyte and polymer chemistry modulate the electronic and structural properties of polymer electrodes during electrochemical (de)doping. These systems are shown to host an ensemble of polarons, in contrast with the single polaron-like character often reported, whose properties vary with the nature of the local environment. Importantly, these polarons serve as reporters of the nanoscale environments in which they reside. We demonstrate that controlling the polymer and electrolyte chemistry regulates the nature of the charge carriers generated upon electrochemical doping and/or exciton dissociation in a photoelectrochemical solar cell: For instance, divalent counterions enable polaron and bipolaron formation at lower reducing potentials, while supporting more bipolaron formation than monovalent counterions. A novel application of NEXAFS reveals insights into charge (de)localization, providing a pathway for future investigation of electron transport mechanisms. Finally, simulations of polymer swelling of an amorphous interphase show that charge formation has a large impact on polymer swelling and ion penetration. These studies deliver insights to enable the control of charge-carrier and ion transport, the rates of electron transfer and catalytic efficiency, device stability, and overall device performance.

14 SOLAR ENERGY↗
DOE OSTI2025

A Greening Future Elevates Flash Drought Risk in Northern Mid‐to‐High Latitudes

Flash droughts have become a growing concern, as they can emerge rapidly and increase the risk of crop failure. Although past studies have investigated the meteorological drivers and future changes of flash drought, why flash drought is more frequent over humid and vegetated regions remains underexplored. This study delves further into the mechanism by which vegetation regulates flash drought and its future change using observations from multiple data sets and large ensemble simulations from three Earth system models. On an interannual timescale, both observations and simulations show robust increases in flash drought frequency and a higher flash-to-sub-seasonal drought ratio during spring or antecedent conditions with dense vegetation, supporting the important role of vegetation in flash drought occurrence, especially in the northern mid-to-high latitudes. In the latter regions, the large ensemble simulations show robust increases in flash drought (e.g., 67% and 46% increases in Eastern U.S. and North Asia in 2050–2100 relative to 1950–2000 under the high emission scenario), where the growing season is lengthening. Although greening might suggest reduced drought stress, it drives precipitation-soil moisture-evapotranspiration decoupling by increasing evapotranspiration partitioning to transpiration. As transpiration can access deep soil water through the plant root system, its increased portion can weaken the constraints of concurrent precipitation on evapotranspiration, thus accelerating soil moisture depletion under high evaporative demand, driving a slow-to-rapid drought transition. How vegetation regulates flash drought by regulating surface moisture budget is supported by observations and simulations. Although warming supports early planting, agriculture may increasingly be threatened by surging flash drought risk.

Drought↗
DOE OSTI2025

Multi-Agent Hierarchical Deep Reinforcement Learning for HVAC Control With Flexible DERs

As electricity consumption in commercial and residential buildings continues to rise, reducing energy costs presents an increasing challenge. Heating, ventilating, and air-conditioning (HVAC) systems, which typically account for 40%-50% of a building's energy use, are prime targets for energy savings. Intelligent control of HVAC temperature through the exploitation of HVAC load flexibility brings significant potential to reduce energy consumption and electricity expenses. The nonlinear models of HVAC systems challenge traditional control methods, while the uncertainty introduced by HVAC load flexibility complicates distributed energy resource (DER) management using conventional optimal dispatch techniques. In response to these challenges, we propose a hierarchical multi-agent deep reinforcement learning (DRL) approach. The lower-level agents focus on balancing comfort and energy conservation, while the upper-level DRL agents optimize the use of DERs to reduce peak demand based on the control outcomes of the HVAC by the lower-level agents. Here, in the upper-level agents, we incorporate a multi-agent structure based on ensemble learning, which acts based on historical and current data without relying on precise load forecasting to address the delayed rewarding issue in DRL. This allows for the effective reduction of energy costs. The proposed method is tested using a real-world microgrid comprising 413 buildings in Southern California, and the results demonstrate that our approach can significantly reduce overall electricity bills while ensuring the comfort of consumers and residents.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2026

Strong Lens Discoveries in DESI Legacy Imaging Surveys DR10 with Two Deep Learning Architectures

Abstract We have conducted a search for strong gravitational lensing systems in the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Surveys Data Release 10 (DR10). This paper is the fourth in a series of searches. This is the first catalog of lens candidates covering nearly the entirety of the extragalactic sky south of declination δ ≈ +32 ∘ , all observed by DECam, covering ∼14,000 deg 2 . We impose a z -band magnitude cut of <20 in AB magnitude. We deploy a residual neural network and EfficientNet as an ensemble trained on a compilation of known lensing systems and high-grade candidates as well as nonlenses in the same footprint. The predictions from these two base models are aggregated using a meta-learner. After applying our ensemble to the survey data, we exclude known candidates and systems, and use our own visual inspection portal to rank images in the top 0.01 percentile of all neural network recommendations. We have found 811 lens candidates, five of which are confirmed through Euclid Quick Data Release (Q1). These include 484 new candidates in the Legacy Surveys DR9 footprint, all parts of which have been searched for strong lenses at least once before, either by our group or others. Combining the discoveries from this work with those from the first three papers in this series (335, 1210, and 1512), we have discovered a total of 3868 new candidates in the DESI Legacy Surveys.

Inchausti, Jose Carlos [University of San Francisc↗
DOE OSTI2024

DLSIA: Deep Learning for Scientific Image Analysis

DLSIA (Deep Learning for Scientific Image Analysis) is a Python-based machine learning library that empowers scientists and researchers across diverse scientific domains with a range of customizable convolutional neural network (CNN) architectures for a wide variety of tasks in image analysis to be used in downstream data processing. DLSIA features easy-to-use architectures, such as autoencoders, tunable U-Nets and parameter-lean mixed-scale dense networks (MSDNets). Additionally, this article introduces sparse mixed-scale networks (SMSNets), generated using random graphs, sparse connections and dilated convolutions connecting different length scales. For verification, several DLSIA-instantiated networks and training scripts are employed in multiple applications, including inpainting for X-ray scattering data using U-Nets and MSDNets, segmenting 3D fibers in X-ray tomographic reconstructions of concrete using an ensemble of SMSNets, and leveraging autoencoder latent spaces for data compression and clustering. As experimental data continue to grow in scale and complexity, DLSIA provides accessible CNN construction and abstracts CNN complexities, allowing scientists to tailor their machine learning approaches, accelerate discoveries, foster interdisciplinary collaboration and advance research in scientific image analysis.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Evaluating Deception Detection Model Robustness To Linguistic Variation

With the increasing use of automated, machine learning-driven tools and the downstream impact that algorithmic judgements can have, it is critical to develop models that are robust to evolving or manipulated inputs. Evaluating the reliability of multimodal models across linguistic variations to understand model susceptibility to intentional linguistic adversarial attacks as well as natural linguistic variations is essential in this pursuit. We present extensive analysis of model robustness and susceptibility to linguistic variations in the setting of deceptive news detection, a difficult classification task that is an increasingly important problem to solve with the impact of misinformation spread online. We evaluate the effectiveness of incorporating adversarial defense strategies and measure model susceptibility to state-of-the-art adversarial attacks using two types of linguistic attacks — character and word perturbations. We consider two multiclass prediction tasks — a 3-way classification of tweets as trustworthy, propaganda, or disinformation; and a 4-way classification as clickbait, hoax, satire, or conspiracy — and compare the performance of three embeddings that have been state-of-the-art for several NLP tasks — GloVe, ELMo, and BERT — to highlight consistent trends in susceptibility, high confidence misclassifications, and high impact failures. We find that character or mixed ensemble models are the most effective defense mechanisms and that character perturbations are a more effective attack than word perturbations for deception classification.

adversarial evaluation↗
DOE OSTI2026

Single-Particle Insights Into the Electronic Structure and Enhanced Stability of CsPbBr 3 /FAPbBr 3 Core/Crown Nanoplatelets at the Nanometer Scale

Colloidal perovskite nanoplatelets (NPLs), despite their exceptional optoelectronic properties, face significant challenges due to their intrinsic instability and trap-assisted non-radiative recombination. Although many studies employing surface engineering, such as selecting ligands or coating to passivate defects, have demonstrated improved optoelectronic properties, these enhancements are typically inferred from bulk-ensemble measurements; the effects at the single-particle level remain elusive. Here, we conduct single-particle-level studies using scanning tunneling spectroscopy (STS) on a unique core–crown system, CsPbBr 3 @FAPbBr 3 NPLs, where the lateral surfaces of the CsPbBr 3 core are coated with an FAPbBr 3 crown. Experimental density-of-states (DOS) analysis reveals a 47% reduction in deep-trap states in core-crown NPLs compared to core-only NPLs, consistent with nearly two-fold enhancements in photoluminescence quantum yields. Progressive I–V sweep measurements demonstrate superior electrical stability in core-crown NPLs, preserving band structure with minimal degradation, while core-only NPLs exhibit rapid bandgap shrinkage and trap formation. Density functional theory (DFT) calculations indicate that FA incorporation distorts Pb octahedral lattice, widening the bandgap. This study elucidates how surface engineering modulates charge localization, passivates defects, and enhances stability at the single-particle level. Moreover, by uncovering bias-induced trap states in single unpassivated NPLs, this study establishes a precise, robust approach for characterizing emerging perovskite nanocrystals.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2022

Cosmological inference from an emulator based halo model. II. Joint analysis of galaxy-galaxy weak lensing and galaxy clustering from HSC-Y1 and SDSS

Here, we present high-fidelity cosmology results from a blinded joint analysis of galaxy-galaxy weak lensing (ΔΣ) and projected galaxy clustering (w p ) measured from the Hyper Suprime-Cam Year-1 (HSC-Y1) data and spectroscopic Sloan Digital Sky Survey (SDSS) galaxy catalogs in the redshift range 0.15 < z < 0.7. We define luminosity-limited samples of SDSS galaxies to serve as the tracers of w p in three spectroscopic redshift bins, and as the lens samples for ΔΣ. For the ΔΣ measurements, we select a single sample of 4×10 6 source galaxies over 140 deg 2 from HSC-Y1 with photometric redshifts (photo z) greater than 0.75, enabling a better handle of photo- z errors by comparing the ΔΣ amplitudes for the three lens redshift bins. The deep, high-quality HSC-Y1 data enable significant detections of the ΔΣ signals, with integrated signal-to-noise ratio S/N ~ 15 in the range 3 ≤ R/[h –1 Mpc] ≤ 30 for the three lens samples, despite the small area coverage. For cosmological parameter inference, we use an input galaxy-halo connection model built on the dark emulator package (which uses an ensemble set of high-resolution N-body simulations and enables fast, accurate computation of the clustering observables) with a halo occupation distribution that includes nuisance parameters to marginalize over modeling uncertainties. We model the ΔΣ and wp measurements on scales from R≃3 and 2h –1 Mpc , respectively, up to 30 h –1 Mpc (therefore excluding the baryon acoustic oscillations information) assuming a flat Λ CDM cosmology, marginalizing over about 20 nuisance parameters and demonstrating the robustness of our results to them. With various tests using mock catalogs described in Miyatake et al. [preceding paper, Phys. Rev. D 106, 083519 (2022)], we show that any bias in the clustering amplitude S 8 ≡ σ 8 (Ω m /0.3) 0.5 due to uncertainties in the galaxy-halo connection is less than ~ 50 % of the statistical uncertainty of S 8 , unless the assembly biaseffect is unexpectedly large. Our best-fit models have S 8 = 0.795$_{-0.042}^{+0.049}$ (mode and 68% credible interval) for the flat Λ CDM model; we find tighter constraints on the quantity S 8 (α = 0.17)≡σ 8 (Ω m /0.3) 0.17 = 0.745$_{-0.031}^{+0.039}$.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Structures of Fission Yeast Inositol Pyrophosphate Kinase Asp1 in Ligand-Free, Substrate-Bound, and Product-Bound States

Expression of the fission yeast Schizosaccharomyces pombe phosphate regulon is sensitive to the intracellular level of the inositol pyrophosphate signaling molecule 1,5-IP 8 . IP 8 dynamics are determined by Asp1, a bifunctional enzyme consisting of an N-terminal kinase domain and a C-terminal pyrophosphatase domain that catalyze IP 8 synthesis and catabolism, respectively. Here, we report structures of the Asp1 kinase domain, crystallized with two protomers in the asymmetric unit, one of which was complexed with ligands (ADPNP, ADP, or ATP; Mg 2+ or Mn 2+ ; IP 6 , 5-IP 7 , or 1,5-IP 8 ) and the other which was ligand-free. The ligand-free enzyme adopts an “open” conformation that allows ingress of substrates and egress of products. ADPNP, ADP, and ATP and associated metal ions occupy a deep phospho-donor pocket in the active site. IP 6 or 5-IP 7 engagement above the nucleotide favors adoption of a “closed” conformation, in which surface protein segments undergo movement and a disordered-to-ordered transition to form an inositol polyphosphate-binding site. In a structure mimetic of the kinase Michaelis complex, the anionic 5-IP 7 phosphates are encaged by an ensemble of nine cationic amino acids: Lys43, Arg223, Lys224, Lys260, Arg274, Arg285, Lys290, Arg293, and Lys341. Alanine mutagenesis of amino acids that contact the adenosine nucleoside of the ATP donor underscored the contributions of Asp258 interaction with the ribose 3'-OH and of Glu248 with adenine-N 6 . Changing Glu248 to Gln elicited a gain of function whereby the kinase became adept at using GTP as phosphate donor. Wild-type Asp1 kinase can utilize N 6 -benzyl-ATP as phosphate donor.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

A Deep Learning-Driven Sampling Technique to Explore the Phase Space of an RNA Stem-Loop

The folding and unfolding of RNA stem-loops are critical biological processes; however, their computational studies are often hampered by the ruggedness of their folding landscape, necessitating long simulation times at the atomistic scale. Here, we adapted DeepDriveMD (DDMD), an advanced deep learning-driven sampling technique originally developed for protein folding, to address the challenges of RNA stem-loop folding. Although tempering- and order parameter-based techniques are commonly used for similar rare-event problems, the computational costs or the need for a priori knowledge about the system often present a challenge in their effective use. DDMD overcomes these challenges by adaptively learning from an ensemble of running MD simulations using generic contact maps as the raw input. DeepDriveMD enables on-the-fly learning of a low-dimensional latent representation and guides the simulation toward the undersampled regions while optimizing the resources to explore the relevant parts of the phase space. We showed that DDMD estimates the free energy landscape of the RNA stem-loop reasonably well at room temperature. Our simulation framework runs at a constant temperature without external biasing potential, hence preserving the information on transition rates, with a computational cost much lower than that of the simulations performed with external biasing potentials. Here, we also introduced a reweighting strategy for obtaining unbiased free energy surfaces and presented a qualitative analysis of the latent space. This analysis showed that the latent space captures the relevant slow degrees of freedom for the RNA folding problem of interest. Finally, throughout the manuscript, we outlined how different parameters are selected and optimized to adapt DDMD for this system. We believe this compendium of decision-making processes will help new users adapt this technique for the rare-event sampling problems of their interest.

Gupta, Ayush↗
DOE OSTI2021

Unleashing the Power of Industrial Big Data through Scalable Manual Labeling

Big Data plays a central role in the remarkable results achieved by Machine Learning (ML) and especially Deep Learning (DL) in the recent years. However, the difficulty in obtaining a reasonable amount of labeled samples limits ML/DL application in various domains, including industrial equipment and system monitoring. In this paper the need for methods that turn manual labeling into a scalable process is highlighted. A real world problem is analyzed for which weak supervision methods, successfully employed in other domains, did not produce acceptable results. An alternative approach based on clustering ensembles is described and tested, achieving good performance.

Paes Leao, Bruno↗
DOE OSTI2025

Operator learning for energy-efficient building ventilation control with computational fluid dynamics simulation of a real-world classroom

Energy-efficient ventilation control plays an important role in reducing building energy consumption while ensuring occupant health and comfort. While Computational Fluid Dynamics (CFD) simulations provide detailed and physically accurate representations of indoor airflow, their high computational cost limits their use in real-time building control. In this work, we present a neural operator learning framework that combines the physical accuracy of CFD with the computational efficiency of machine learning to enable building ventilation control with the high-fidelity fluid dynamics models. Our method jointly optimizes the airflow supply rates and vent angles to reduce energy use and adhere to air quality constraints. We train an ensemble of neural operator transformer models to learn the mapping from building control actions to airflow fields using high-resolution CFD data. This learned neural operator is then embedded in an optimization-based control framework for building ventilation control. Experimental results show that our approach achieves significant energy savings compared to maximum airflow rate control, rule-based control, as well as data-driven control methods using spatially averaged CO 2 prediction and deep learning–based reduced-order models, while consistently maintaining safe indoor air quality. These results highlight the practicality and scalability of our method in maintaining energy efficiency and indoor air quality in real-world buildings.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Event Classifications on DNE2 Main Experiment Data using a Convolutional Neural Network Ensemble

The Dynamic Networks (DN) Experiment for FY24 (DNE2) is an experiment within DN with the goal of quantitatively evaluating the effectiveness of solutions developed so far by various researchers under the Low Yield Nuclear Monitoring (LYNM) program using a shared set of metrics and datasets. A key component of this experiment is the mimicking of a signature processing pipeline, and comparing currently accepted and standard-use processing methods to more state-of-the-art processes developed under DN. In this work, we focus specifically on the Event Characterization (EC) Focus Area (FA) of the pipeline, where a seismic event’s magnitude, yield and class are identified. We use Deep Learning (DL) to classify the type of events being processed as either earthquakes (EQs) or explosions (EXs) for three iterations of experiment datasets. The model is noticeably more confident and accurate in classifying explosions than earthquakes, reflecting a known shortcoming of the model, that being of a bias towards predicting explosions over earthquakes in the west coast due to training data biases.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Statistical characterization of experimental magnetized liner inertial fusion stagnation images using deep-learning-based fuel–background segmentation

Significant variety is observed in spherical crystal x-ray imager (SCXI) data for the stagnated fuel–liner system created in Magnetized Liner Inertial Fusion (MagLIF) experiments conducted at the Sandia National Laboratories Z-facility. As a result, image analysis tasks involving, e.g., region-of-interest selection (i.e. segmentation), background subtraction and image registration have generally required tedious manual treatment leading to increased risk of irreproducibility, lack of uncertainty quantification and smaller-scale studies using only a fraction of available data. We present a convolutional neural network (CNN)-based pipeline to automate much of the image processing workflow. This tool enabled batch preprocessing of an ensemble of N scans = 139 SCXI images across N exp = 67 different experiments for subsequent study. The pipeline begins by segmenting images into the stagnated fuel and background using a CNN trained on synthetic images generated from a geometric model of a physical three-dimensional plasma. The resulting segmentation allows for a rules-based registration. Our approach flexibly handles rarely occurring artifacts through minimal user input and avoids the need for extensive hand labelling and augmentation of our experimental dataset that would be needed to train an end-to-end pipeline. Here we also fit background pixels using low-degree polynomials, and perform a statistical assessment of the background and noise properties over the entire image database. Our results provide a guide for choices made in statistical inference models using stagnation image data and can be applied in the generation of synthetic datasets with realistic choices of noise statistics and background models used for machine learning tasks in MagLIF data analysis. We anticipate that the method may be readily extended to automate other MagLIF stagnation imaging applications.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

DiffESM: Conditional Emulation of Temperature and Precipitation in Earth System Models With 3D Diffusion Models

Earth system models (ESMs) are essential for understanding the interaction between human activities and the Earth's climate. However, the computational demands of ESMs often limit the number of simulations that can be run, hindering the robust analysis of risks associated with extreme weather events. While low-cost climate emulators have emerged as an alternative to emulate ESMs and enable rapid analysis of future climate, many of these emulators only provide output on at most a monthly frequency. This temporal resolution is insufficient for analyzing events that require daily characterization, such as heat waves or heavy precipitation. We propose using diffusion models, a class of generative deep learning models, to effectively downscale ESM output from a monthly to a daily frequency. Trained on a handful of ESM realizations, reflecting a wide range of radiative forcings, our DiffESM model takes monthly mean precipitation or temperature as input, and is capable of producing daily values with statistical characteristics close to ESM output. Combined with a low-cost emulator providing monthly means, this approach requires only a small fraction of the computational resources needed to run a large ensemble. We evaluate model behavior using a number of extreme metrics, showing that DiffESM closely matches the spatio-temporal behavior of the ESM output it emulates in terms of the frequency and spatial characteristics of phenomena such as heat waves, dry spells, or rainfall intensity.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Early-time measure in eternal inflation

In a situation like eternal inflation, where our data is replicated at infinitely-many other space-time events, it is necessary to make a prior assumption about our location to extract predictions. The principle of mediocrity entails that we live at asymptotic late times, when the occupational probabilities of vacua has settled to a near-equilibrium distribution. In this paper we further develop the idea that we instead exist during the approach to equilibrium, much earlier than the exponentially-long mixing time. In this case we are most likely to reside in vacua that are easily accessed dynamically. Using first-passage statistics, we prove that vacua that maximize their space-time volume at early times have: 1. maximal ever-hitting probability; 2. minimal mean first-passage time; and 3. minimal decay rate. These requirements are succinctly captured by an early-time measure. The idea that we live at early times is a predictive guiding principle, with many phenomenological implications. First, our vacuum should lie deep in a funneled region, akin to folding energy landscapes of proteins. Second, optimal landscape regions are characterized by relatively short-lived vacua, with lifetime of order the de Sitter Page time. For our vacuum, this lifetime is ~ 10 130 years, which is consistent with the Standard Model estimate due to Higgs metastability. Third, the measure favors vacua with small, positive vacuum energy. This can address the cosmological constant problem, provided there are sufficiently many vacua in the entire ensemble of funnels. As a concrete example, we study the Bousso-Polchinski lattice of flux vacua, and find that the early-time measure favors lattices with the fewest number of flux dimensions. This favors compactifications with a large hierarchy between the lightest modulus and all other Kähler and complex structure moduli.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Enhancing Automotive Intrusion Detection Through Multi-Modal Fusion: A CAN FD-LiDAR Approach

As vehicles become smarter and more autonomous, they increasingly depend on advanced sensors and communication technologies to operate securely. However, such growing dependence on technology—whether it’s CAN (Controller Area Network) for internal communication or LiDAR (Light Detection and Ranging) for sensing the world around them—also expands the attack surface for the types of cyber attacks. Traditional intrusion detection systems (IDS) typically monitor these systems in isolation, limiting their ability to detect sophisticated, crosssystem attacks. To address this, we propose a multi-modal fusion approach that combines real-world CAN FD signals (from the HCRL dataset) with LiDAR features (from the nuScenes dataset) to enhance attack detection. Our method employs a twostage ensemble approach. Calibrated XGBoost and LightGBM models initially process CAN FD (Fuzzing Data) and LiDAR data independently, detecting timing anomalies and space abnormalities. They are subsequently logarithmically combined with a logistic regression meta-model along with 17 engineered features capturing cross-modal behavior, prediction conflicts, and nonlinear interactions. This approach achieves an AUC of 0.87 and an F1-score of 0.82, surpassing single-modality baselines and early fusion methods, at merely 2 ms inference latency. Compared with deep learning competitors, it is 3 times more efficient, providing a lightweight, interpretable, and real time solution to automotive cybersecurity.

97 MATHEMATICS AND COMPUTING↗