Time Window Size Research Articles

A benchmark for domain adaptation and generalization in smartphone-based human activity recognition

Human activity recognition (HAR) using smartphone inertial sensors, like accelerometers and gyroscopes, enhances smartphones’ adaptability and user experience. Data distribution from these sensors is affected by several factors including sensor hardware, software, device placement, user demographics, terrain, and more. Most datasets focus on providing variability in user and (sometimes) device placement, limiting domain adaptation and generalization studies. Consequently, models trained on one dataset often perform poorly on others. Despite many publicly available HAR datasets, cross-dataset generalization remains challenging due to data format incompatibilities, such as differences in measurement units, sampling rates, and label encoding. Hence, we introduce the DAGHAR benchmark, a curated collection of datasets for domain adaptation and generalization studies in smartphone-based HAR. We standardized six datasets in terms of accelerometer units, sampling rate, gravity component, activity labels, user partitioning, and time window size, removing trivial biases while preserving intrinsic differences. This enables controlled evaluation of model generalization capabilities. Additionally, we provide baseline performance metrics from state-of-the-art machine learning models, crucial for comprehensive evaluations of generalization in HAR tasks.

Deep learning of many-body observables and quantum information scrambling

Machine learning has shown significant breakthroughs in quantum science, where in particular deep neural networks exhibited remarkable power in modeling quantum many-body systems. Here, we explore how the capacity of data-driven deep neural networks in learning the dynamics of physical observables is correlated with the scrambling of quantum information. We train a neural network to find a mapping from the parameters of a model to the evolution of observables in random quantum circuits for various regimes of quantum scrambling and test its generalization and extrapolation capabilities in applying it to unseen circuits. Our results show that a particular type of recurrent neural network is extremely powerful in generalizing its predictions within the system size and time window that it has been trained on for both, localized and scrambled regimes. These include regimes where classical learning approaches are known to fail in sampling from a representation of the full wave function. Moreover, the considered neural network succeeds in extrapolating its predictions beyond the time window and system size that it has been trained on for models that show localization, but not in scrambled regimes.

An Adaptively Weighted Averaging Method for Regional Time Series Extraction of fMRI-based Brain Decoding.

Brain decoding that classifies cognitive states using the functional fluctuations of the brain can provide insightful information for understanding the brain mechanisms of cognitive functions. Among the common procedures of decoding the brain cognitive states with functional magnetic resonance imaging (fMRI), extracting the time series of each brain region after brain parcellation traditionally averages across the voxels within a brain region. This neglects the spatial information among the voxels and the requirement of extracting information for the downstream tasks. In this study, we propose to use a fully connected neural network that is jointly trained with the brain decoder to perform an adaptively weighted average across the voxels within each brain region. We perform extensive evaluations by cognitive state decoding, manifold learning, and interpretability analysis on the Human Connectome Project (HCP) dataset. The performance comparison of the cognitive state decoding presents an accuracy increase of up to 5% and stable accuracy improvement under different time window sizes, resampling sizes, and training data sizes. The results of manifold learning show that our method presents a considerable separability among cognitive states and basically excludes subject-specific information. The interpretability analysis shows that our method can identify reasonable brain regions corresponding to each cognitive state. Our study would aid the improvement of the basic pipeline of fMRI processing.

Predicting problem-solving success in an office simulation applying N-grams and a random forest to behavioral process data

Predicting students' problem-solving success in computer-based simulations at an early stage allows adaptive educational systems to provide learners with personalized support. In this paper, we predict students' problem-solving success by applying a machine-learning model, the random forest, to produce a binary classification (more vs. less successful students). During a business-related problem scenario that lasted 55 min, early behavioral data (during the first 5, 10, and 20 min) such as mouse clicks and keyboard strokes (approx. 29,800 early-window clickstreams and keystrokes during the first 20 min) of 234 trainees were recorded, mirroring the students' problem-solving behavior. We used the n-gram sequence mining technique, which was originally introduced within the emerging disciplines of natural language processing, text mining, and machine learning and has proven to be effective, particularly in the examination of online behavior. We trained the random forest model with training datasets that included all features (bigrams), as well as selected features (the most predictable bigrams explaining inter-group differences). Our results show that early predictions based on the first 10 and 20 min contained sufficient information to accurately predict problem-solving success, while predictions that are too early (based on the first 5 min) do not. As the size of the initial time window expanded, the classification performance improved. Moreover, the selection of the most predictable features improved the models' performance for all three time intervals. The model that was trained with only selected robust features that occurred in the first 20 min achieved the highest ROC AUC score of almost 0.70. This result falls within the range of accuracy scores observed in similar studies. From the instructor's perspective, predictions help in the early identification of weak students and can provide them with personalized learning prompts. For more successful students, tasks can be enriched adaptively.

Safety evaluation and prediction of takeover performance in automated driving considering drivers’ cognitive load: A driving simulator study

Automated vehicles alleviate the need for driver attention and control. However, a takeover request (TOR) to the driver remains essential for emergencies and safety–critical scenarios beyond automation's capability. Thus, accessing drivers’ safety performance in various TOR scenarios is crucial for conditionally automated driving (SAE L3). However, TOR safety performance is seldom examined concerning drivers' cognitive load, despite its presumed relevance to TOR scenarios and human–machine interaction. Moreover, the adequacy of the time window preceding a TOR, critical for TOR safety prediction, remains inadequately explored. This study aims to assess safety performance across diverse TOR scenarios in Level 3 conditional automation, incorporating drivers’ cognitive load, and predict TOR safety by considering the time window's impact. A driving simulator experiment gathered eye movement and driving behavior data from 37 recruited participants. Participants were instructed to take control of the vehicle from automated driving within a time budget (TB) of 3 s or 7 s in obstacle avoidance (OA) or lane keeping (LK) scenarios while engaging in non-driving-related tasks (NDRTs). Participants’ subjective cognitive load in the different TOR scenarios was scaled using NASA-TLX. Furthermore, safe TOR performance was predicted utilizing Convolutional Neural Networks (CNNs) across different time window sizes preceding a TOR. The results indicate that: 1) cognitive loads in takeover scenarios ranked from highest to lowest as TB = 3s_OA, TB = 3s_LK, TB = 7s_OA, TB = 7s_LK; and the cognitive loads of the NDRTs ranked from highest to lowest as mistake finding, texting, chatting, monitoring; 2) the takeover safety performance in the four scenarios from lowest to highest was TB = 3s_OA, TB = 3s_LK, TB = 7s_OA, TB = 7s_LK; likewise, the takeover safety performance during the four NDRTs ranked from lowest to highest as mistake finding, monitoring, texting, chatting; 3) the time window size before the TORs significantly affected the prediction performance of the model. A 30-second window was recommended as optimal for predicting takeover safety using the CNN model, achieving an average F1 score of 0.8120 and 81.98 % accuracy. This study's findings enhance our comprehension of driving behavior characteristics during TOR and offer valuable insights for detecting driver states in conditionally automated driving contexts.

Forecasting Model of Sustainable Financial Market Trend Based on Neural Network Algorithm

Recent advances in computer technology have resulted in the ongoing gathering of massive amounts of data and information. Because the financial market creates so much real-time data, including transaction records, we have a great potential to get important insights from analysing that data, especially in the banking industry. As a result, the goal of this work is to use the financial data that is now accessible to create a unique stock market prediction model. We employ the deep learning approach because of its exceptional capacity to learn from large datasets. This study proposes a hybrid approach that integrates the Archimedes optimisation algorithm (AOA) with a long short-term memory (LSTM) network. So far, heuristic-based trial and error has been extensively used to estimate the temporal window size and architectural components of long short-term memory networks. This paper investigates the temporal properties of financial market data by providing a systematic way to selecting the topology and time window size for the LSTM network. The experimental results demonstrate that the hybrid LSTM network and AOA model outperform the benchmark model.

A dedicated in-beam PET system with a modular dual-head for radiotherapy imaging in HIMM

This paper presents a heavy-ion in-beam PET, referred to as ibPET with a specifically custom designed data acquisition system, which was constructed and tested at the Heavy Ion Medical Machine (HIMM) beam-line, HICTC in Wuwei. The data acquisition system of ibPET has undergone several upgrades during the years, And its data processing capability now exceeds one million events per second per channel. To achieve optimal performance of the ibPET under existing conditions, we tested coincident events and noise equivalent count (NEC) under different time coincidence windows.The results show that the signal-to-noise ratio of the reconstructed image data reaches its peak when the time window is around 3 ns, indicating that the ibPET performance is optimal at this point. In order to determine the size of the time window, we measured the system's time resolution FWHM to be 1.63 ns, and ultimately chose a time window size of twice the time resolution FWHM, which is 3.25 ns. Meanwhile, we demonstrate the capability of our ibPET to monitor the dose range with various irradiation. The results demonstrate that the system exhibits the expected exponential decay in counting rate after irradiation, and accurately provides penetration depth positions at different energy levels.

Optimal Service Time Windows

Because customers must usually arrange their schedules to be present for home services, they desire an accurate estimate of when the service will take place. However, even when firms quote large service time windows, they are often missed, leading to customer dissatisfaction. Wide time windows and frequent failures occur because time windows must be communicated to customers in the face of several uncertainties: future customer requests are unknown, final service plans are not yet determined, and when fulfillment is outsourced to a third party, the firm has limited control over routing procedures and eventual fulfillment times. Even when routing is performed in-house, time windows often do not receive explicit consideration. In this paper, we show how companies can communicate reliable and narrow time windows to customers in the face of arrival time uncertainty when time window decisions are decoupled from routing procedures. Under assumptions on the shape of arrival time distributions, our main result characterizes the optimal policy, identifying structure that reduces a high-dimensional stochastic nonlinear optimization problem to a root-finding problem in one dimension. The result inspires a practice-ready heuristic for the more general case. Relative to the industry standard of communicating uniform time windows to all customers, and to other policies applied in practice, our method of quoting customer-specific time windows yields a substantial increase in customer convenience without sacrificing reliability of service. Our results show that time windows should be tailored to individual customers, time window sizes should be proportional to the service level, larger time windows should be assigned to earlier requests and smaller time windows to later requests, larger time windows should be assigned to customers further from the depot of operation and smaller time windows to closer customers, high quality time windows can be identified even with limited data, and cost savings afforded by routing efficiency should be measured against potential losses to customer convenience. Funding: M. W. Ulmer’s work is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) Emmy Noether Programme, [project 444657906]. Supplemental Material: The e-companion is available at https://doi.org/10.1287/trsc.2023.0004 .

Bearing early fault identification and condition monitoring based on dynamic difference index

Bearing condition monitoring is essential for early fault detection and early warning of large equipment, and signal processing techniques are frequently used to analyse nonlinear and nonstationary sequences. Cross-correlation integral is implemented in bearing condition monitoring because it can analyse the non-stationarity of time series in dynamic systems. This paper proposes a dynamic difference index (DDI) due to the difficulty of determining the threshold in the cross-correlation integral and the roughness and operation caused by sequence similarity of 0 or 1. It is a measure of the similarity between the fuzzy autocorrelation integral of a portion of a time series and the cross-correlation between that portion and other portions of the same time series, and it is used to determine the stationarity of the time series. When bearings begin to degrade or develop structural defects, the DDI changes dramatically. The XJTU-SY dataset and IMS Bearing Data were utilised for algorithm validation. First, the algorithm’s efficacy was demonstrated by optimising the effects of various thresholds, distance measures, and time window sizes on DDI and computational efficiency. Secondly, the comparison with common methods and state of the art shows the superiority of the algorithm in detecting early bearing faults, and reveals its appropriate use in practical engineering applications.

CQ + Training: Minimizing Accuracy Loss in Conversion from Convolutional Neural Networks to Spiking Neural Networks.

Spiking neural networks (SNNs) are attractive for energy-constrained use-cases due to their binarized activation, eliminating the need for weight multiplication. However, its lag in accuracy compared to traditional convolutional network networks (CNNs) has limited its deployment. In this paper, we propose CQ+ training (extended "clamped" and "quantized" training), an SNN-compatible CNN training algorithm that achieves state-of-the-art accuracy for both CIFAR-10 and CIFAR-100 datasets. Using a 7-layer modified VGG model (VGG-*), we achieved 95.06% accuracy on the CIFAR-10 dataset for equivalent SNNs. The accuracy drop from converting the CNN solution to an SNN is only 0.09% when using a time step of 600. To reduce the latency, we propose a parameterized input encoding method and a threshold training method, which further reduces the time window size to 64 while still achieving an accuracy of 94.09%. For the CIFAR-100 dataset, we achieved an accuracy of 77.27% using the same VGG-* structure and a time window of 500. We also demonstrate the transformation of popular CNNs, including ResNet (basic, bottleneck, and shortcut block), MobileNet v1/2, and Densenet, to SNNs with near-zero conversion accuracy loss and a time window size smaller than 60. The framework was developed in PyTorch and is publicly available.

SNPU: An Energy-Efficient Spike Domain Deep-Neural-Network Processor With Two-Step Spike Encoding and Shift-and-Accumulation Unit

In this article, an energy-efficient spike domain deep-neural-network processor (SNPU) is proposed. Recently, many sparsity-aware processors were proposed to increase energy efficiency. However, because the activation of the real-world network was not high compared to the ideal condition, they were unable to completely employ integrated zero skipping logic. In addition, they employed weight sparsity by pruning, but their zero skipping logic was designed to perform best only under specific sparsity condition. We adopt two-step encoding (neural and computing encoding) and an omni-sparsity handling unit to solve these issues. The input neural encoding eliminates pseudo-random number generators (PRNGs) for the previous encoding, and the output neural encoding reduces redundant memory access. As a result, weight can be reused by a delayed factor, while input spike encoding consumes 99.4% less power compared to the Poisson encoder. Also, SNPU uses computing encoding with multi-level encoding and spiketrain decomposition. The multi-level encoding reduces the time-window size by merging several spikes, and SNPU uses a single shift-and-accumulation (SAC) rather than several accumulations, reducing 43.5% of operations compared to single-level encoding. In addition, the spike-train decomposition is proposed to decrease the number of SACs by computation reuse, resulting in a reduction of 71.8% operations compared to previous spiking neural-network (SNN) processing, which cannot reuse computation result. As a result, the computing encoding reduces the operations by 83.3% compared to SNPU without multi-level encoding and spike-train decomposition. Moreover, SAC units in SNN PE process multi-level spikes while consuming 80.3% less power than accumulator arrays that do not use the two-step encoding. Also, the neuron link and spike-train allocator handle variable time-window sizes, and they can increase PE utilization of SNPU by 23.1%. To sum it up, SNPU shows the state-of-the-art energy efficiency of 13.7 TOPS/W for object detection and 31.5 TOPS/W for ImageNet classification. Moreover, SNPU shows the state-of-the-art power consumption of 63.2 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu$</tex-math> </inline-formula> W for face recognition.

The Teddy tool v1.1: temporal disaggregation of daily climate model data for climate impact analysis

Abstract. Climate models provide the required input data for global or regional climate impact analysis in temporally aggregated form, often in daily resolution to save space on data servers. Today, many impact models work with daily data; however, sub-daily climate information is becoming increasingly important for more and more models from different sectors, such as the agricultural, water, and energy sectors. Therefore, the open-source Teddy tool (temporal disaggregation of daily climate model data) has been developed to disaggregate (temporally downscale) daily climate data to sub-daily hourly values. Here, we describe and validate the temporal disaggregation, which is based on the choice of daily climate analogues. In this study, we apply the Teddy tool to disaggregate bias-corrected climate model data from the Coupled Model Intercomparison Project Phase 6 (CMIP6). We choose to disaggregate temperature, precipitation, humidity, longwave radiation, shortwave radiation, surface pressure, and wind speed. As a reference, globally available bias-corrected hourly reanalysis WFDE5 (WATCH Forcing Data methodology applied to ERA5) data from 1980–2019 are used to take specific local and seasonal features of the empirical diurnal profiles into account. For a given location and day within the climate model data, the Teddy tool screens the reference data set to find the most similar meteorological day based on rank statistics. The diurnal profile of the reference data is then applied on the climate model. The physical dependency between variables is preserved, since the diurnal profile of all variables is taken from the same, most similar meteorological day of the historical reanalysis dataset. Mass and energy are strictly preserved by the Teddy tool to exactly reproduce the daily values from the climate models. For evaluation, we aggregate the hourly WFDE5 data to daily values and apply the Teddy tool for disaggregation. Thereby, we compare the original hourly data with the data disaggregated by Teddy. We perform a sensitivity analysis of different time window sizes used for finding the most similar meteorological day in the past. In addition, we perform a cross-validation and autocorrelation analysis for 30 globally distributed samples around the world that represent different climate zones. The validation shows that Teddy is able to reproduce historical diurnal courses with high correlations &gt;0.9 for all variables, except for wind speed (&gt;0.75) and precipitation (&gt;0.5). We discuss the limitations of the method regarding the reproduction of precipitation extremes, interday connectivity, and disaggregation of end-of-century projections with strong warming. Depending on the use case, sub-daily data provided by the Teddy tool could make climate impact assessments more robust and reliable.

Emotion Recognition Through Combining EEG and EOG Over Relevant Channels With Optimal Windowing

For dimensional emotion recognition, electroencephalography (EEG) signals and electrooculogram (EOG) signals are often combined to improve the performance of classifiers, as each of them provides complementary features to the other. In this article, we combine the EEG signal on the relevant channels with the EOG signal to boost the recognition accuracy. We first explore the mutual information (MI) of all EEG channels and only select emotion-related channels, i.e., channels with more MI are retained, since the emotion recognition performance can be degraded by the interference between uncorrelated channels, while the computational complexity is significant if all EEG channels are used for recognition. While the optimal lengths of EEG and EOG signals for emotion recognition are still uncertain, we systematically investigate the effects of time-window size on emotion recognition. This strategy not only increases the number of training samples, but also reduces the feature redundancy. At this stage, we not only extract multiple statistical features but also employ the increment entropy to find abrupt changes in EEG signals. The experimental results show that 13 out of 32 EEG channels were selected by the proposed channel selection algorithm, and these selected channels can already produce accurate emotion predictions. We found that using optimal time-windows to split EEG and EOG signals into several thin slices and then combine them can further enhance the emotion recognition performance, where the time-windows of 4, 5, 6, and 10 s allow the combined signals to achieve very high accuracy.

Eliminating Mapping Error of Link Quality Prediction for Low-Power Wireless Networks

The size of time windows used by low power wireless protocols to compute packet reception ratio (PRR) directly affects the accuracy and agility of link quality prediction. Using small time windows would make the prediction more agile, but its accuracy decreases seriously. To improve the accuracy while maintaining agility, many existing methods generally predict the physical layer parameters within small time windows first and then calculate PRR using mapping models of such parameters and PRR. However, due to the propagation characteristics of wireless signal, these mapping models usually introduce large errors. The mapping error will superimpose on the prediction error of physical layer parameters, which would inevitably degrade the accuracy and reliability of PRR prediction. This paper proposes to eliminate the step of mapping from physical layer parameters to PRR and directly use the historical series of PRR computed within small time windows as input parameters to build prediction model. Then, the PRR within large time windows could be predicted. This effectively avoids the errors introduced by using mapping models while not sacrificing the agility. To verify the proposed method, several machine learning algorithms were chosen to implement the prediction model. Compared with similar methods based on mapping models, the proposed one could achieve higher accuracy under different size of time windows. Specifically, under large time windows which are commonly used to describe PRR in practice, its accuracy is much higher. More importantly, agility of the proposed method is basically equivalent or even superior to existing ones.

Community evolution prediction based on a self-adaptive timeframe in social networks

The prediction of community evolution events in dynamic social networks is of great importance for network security alerts. Currently, most of the common prediction models use a fixed timeframe division strategy to divide the dynamic social network, such as the disjoint timeframe division strategy and overlapping timeframe division strategy. However, these frameworks cannot change once they are selected, and are not suitable for the prediction of network evolution with strong variability in real-world applications. In this paper, a new community evolution model is developed from the perspective of the universality of the timeframe, and a new optimized timeframe partitioning algorithm is proposed. Compared with the traditional fixed timeframe partitioning algorithm, this method adaptively adjusts the size and number of time windows according to the information fluctuations of the specific network at an acceptable extra computational cost. Based on the analysis of several real-world networks, we found that the proposed self-adaptive timeframe partitioning algorithm improved the quality of community tracking of the network and ensured the accuracy of prediction events.

An XGBoost approach to detect driver visual distraction based on vehicle dynamics

Objectives Distracted driving such as reading phone messages during driving is risky, as it increases the probability of severe crashes. This study proposes an XGBoost model for visual distraction detection based on vehicle dynamics data from a driving simulation study. Methods A simulated driving experiment involving thirty-six drivers was launched. We obtained the vehicle dynamics parameters required for the model using the time window and fast Fourier transform methods, totaling 26 items. Meanwhile, the effects of varied time window sizes (1–7 s) and amount of input indications on model performance were studied. Results By conducting a comparative analysis, it has been determined that the ideal time window size is 5 s. Additionally, the optimal number of input indicators was found to be 23. The XGBoost model for distinguishing distractions achieved an accuracy rate of 85.68%, a precision rate of 85.83%, a recall rate of 83.85%, an F1 score of 84.82%, and an AUC value of 0.9319, which were higher than SVM and RF. The gain-based feature rank demonstrated that the standard deviation of vehicle sideslip rate and the mean amplitude of the 0–1 Hz spectrum component of the steering wheel angle were more crucial than other features. Conclusions The research results indicate that the steering wheel angle and vehicle sideslip angle may be more conducive to identifying distractions. This XGBoost model could potentially be applied in advanced driving assistant systems (ADAS) to warn driver and reduce cellphone involved distracted driving.

Relationship between braking intensity and driver heart rate as a function of the size of the measurement window and its position

Heart Rate (HR) is often used to capture the driver's mental workload and stress. Usually, during a predefined time window (e.g., 10 s, 30 s, or longer). In addition, the center of that time window is commonly placed at the peak of events of interest (e.g., safety events such as hard braking). However, the choices of window size and window position relative to driving events are often not explained in the literature, and conceivably, other choices could have been made for estimating changes in HR. This research tested the effect of window size and position on HR estimates. Participants executed 12 braking events on a closed driving course and were monitored for HR and skin conductivity. We fitted a regression model on HR as a function of driver braking intensity, multiple times (N = 210), each time, we calculated the HR with different window sizes (1–70 s) and three window positions (onset of braking, peak deceleration, and end of braking). We found that the coefficients in the model on HR by braking intensity changed as a function of the window size and its position. Our sample's optimal time window size was different from conventional settings: it was 4-second long and centered at the end of the braking events. Our findings showcase the analysis procedure that we employed and can contribute to research on physiological markers (HR and others) and to the design of driver assistance systems.

Energy efficiency and delay determinacy tradeoff in energy harvesting-powered zero-touch deterministic industrial M2M communications

Data transmission for Industrial Internet of Things (IIoTs) is of the utmost importance, especially in the Industry 5.0 era, where human–machine collaboration is increasingly intensive. On the one hand, as a key characteristic for Industry 5.0, deterministic transmission aims to ensure the data arrive at destination devices accurately, has attracted remarkable attentions. On the other hand, energy efficiency problem is another major concern in Industry 5.0 since the massive Machine-Type Devices (MTDs). Consequently, in this paper, we investigate the tradeoff between energy efficiency and delay determinacy in energy harvesting-powered Zero-Touch Deterministic Industrial Machine-to-Machine (ZT-DI-M2M) communications. In particular, we first derive the probability of the transmission delay falling within a certain time window to characterize the delay determinacy and then figure out the relationship between delay determinacy and the energy efficiency by taking the system control performance into account. After that, in order to model the uncertainty of the stochastic environment, we formulate the problem as a stochastic game by jointly considering transmission power control, frequency spectrum allocation and the selection of base stations. Afterwards, a random graph-based sparse Long Short-Term Memory (LSTM) network is proposed to solve the optimization problem while reducing the computational complexity. Finally, numerical result demonstrates that the proposed sparse LSTM algorithm enables to achieve a specific delay determinacy with a higher energy efficiency as compared to Deep Q-Learning Network (DQN) and Deep Deterministic Policy Gradient (DDPG) algorithms. In addition, we also analysis the influence of sparse coefficient and the size of time window on the relationship between energy efficiency and delay determinacy.

DT-DS: CAN Intrusion Detection with Decision Tree Ensembles

The controller area network (CAN) protocol, used in many modern vehicles for real-time inter-device communications, is known to have cybersecurity vulnerabilities, putting passengers at risk for data exfiltration and control system sabotage. To address this issue, researchers have proposed to utilize security measures based on cryptography and message authentication; unfortunately, such approaches are often too computationally expensive to be deployed in real time on CAN devices. Additionally, they have developed machine learning (ML) techniques to detect anomalies in CAN traffic and thereby prevent attacks. The main disadvantage of existing ML-based techniques is that they either depend on additional computational hardware or they heuristically assume that all communication anomalies are malicious. In this article, we show that tree-based learning ensembles outperform anomaly-based techniques like AutoRegressive Integrated Moving Average (ARIMA) and Z-Score when used to detect attacks that result in increased bus utilization. We evaluated the detection capacity of three tree-based ensembles, Adaboost, gradient boosting, and random forests, and collectively refer to these as DT-DS. We conclude that the decision tree ensemble with Adaboost performs best with an area under curve (AUC) score of 0.999, closely followed by gradient boosting and random forests with 0.997 and 0.991 AUC scores, respectively, when trained using message profiles. We observe that with an increase in the observation window, the DT-DS models present an average AUC score of 0.999, and offer a nearly perfect detection of attacks, at the cost of increased latency in detection of attacked messages. We evaluate the performance of the IDS for Aeronautical Radio, Incorporated– (ARINC) encoded CAN communication traffic in avionic systems, generated using an aerospace testbench, ARINC-825TBv2. The IDS has been evaluated against the active attacks of a state-of-the-art predictive attacker model. Additionally, we observed that the performance of IDS approaches such as ARIMA and Z-Score degrade considerably with a decrease in the size of the observation time window. In contrast, the performance of DT-DS models is consistent, with only an average drop of 0.005 in the AUC score.

Extraction of Information on the Flooding Extent of Agricultural Land in Henan Province Based on Multi-Source Remote Sensing Images and Google Earth Engine

Sudden flood disasters cause serious damage to agricultural production. Rapidly extracting information such as the flooding extent of agricultural land and capturing the influence of flooding on crops provides important guidelines for estimating the flood-affected area, promoting post-disaster farmland restoration, and providing an auxiliary decision-making basis for flood prevention and disaster relief departments. Taking the flood event in Henan and Shanxi Provinces as example, based on the characteristics of the variations in radar data and optical data before and after the disaster, we propose an extent information extraction method for the flood inundation area and the flood-affected area of agricultural land. This method consists of change detection, threshold extraction, and superposition analysis, which weakens the negative impact of the radar data speckle noise and cloud contamination of the optical data on the extraction of the agricultural land flooding to a certain extent. The method was developed based on a flood event in Henan Province and validated in Shanxi Province. The results show that the production of this method have a clear boundary and accurate extent, and the overall precisions of the flood inundation area and flood-affected area extraction are 0.87 and 0.92, respectively. The proposed method combines the advantages of both radar and optical remote sensing data in extracting the specific extents of the flood inundation area and the flood-affected area in large spatial scale. Finally, the impact of time window size to the performance of the method is further analyzed. In the application of the proposed method, the Google Earth Engine (GEE) platform provides a low-cost, fast, and convenient way to extract flood information from remote sensing data. The proposed scheme provides a scientific data basis for restoring production of agricultural land after a flood disaster, as well as for national post-disaster damage assessment and disaster relief decision making.