A Deep Learning Methodology for Predicting Cybersecurity Attacks on the Internet of Things

Information, Год журнала: 2023, Номер 14(10), С. 550 - 550

Опубликована: Окт. 7, 2023

With the increasing severity and frequency of cyberattacks, rapid expansion smart objects intensifies cybersecurity threats. The vast communication traffic data between Internet Things (IoT) devices presents a considerable challenge in defending these from potential security breaches, further exacerbated by presence unbalanced network data. AI technologies, especially machine deep learning, have shown promise detecting addressing threats targeting IoT networks. In this study, we initially leverage learning algorithms for precise extraction essential features realistic-network-traffic BoT-IoT dataset. Subsequently, assess efficacy ten distinct models malware. Our analysis includes two single classifiers (KNN SVM), eight ensemble (e.g., Random Forest, Extra Trees, AdaBoost, LGBM), four architectures (LSTM, GRU, RNN). We also evaluate performance enhancement when integrated with SMOTE (Synthetic Minority Over-sampling Technique) algorithm to counteract imbalanced Notably, CatBoost XGBoost achieved remarkable accuracy rates 98.19% 98.50%, respectively. findings offer insights into ML DL techniques, conjunction balancing such as SMOTE, effectively identify intrusions.

Язык: Английский

---

Dynamic Graph Convolutional Network-Based Prediction of the Urban Grid-Level Taxi Demand–Supply Imbalance Using GPS Trajectories

ISPRS International Journal of Geo-Information, Год журнала: 2024, Номер 13(2), С. 34 - 34

Опубликована: Янв. 24, 2024

The objective imbalance between the taxi supply and demand exists in various areas of city. Accurately predicting this helps companies with dispatching, thereby increasing their profits meeting travel needs residents. application Graph Convolutional Networks (GCNs) traffic forecasting has inspired development a spatial–temporal model for grid-level prediction demand–supply imbalance. However, GCN models conventionally capture only static inter-grid correlation features. This research aims to address dynamic influences caused by mobility variations other transportation modes on dynamics grids. To achieve this, we employ trajectory data develop that incorporates Gated Recurrent Units (GRUs) predict imbalances. captures neighboring grids spatial dimension. It also identifies trends periodic changes temporal validation model, using from Shenzhen city, indicates superior performance compared classical time-series models. An ablation study is conducted analyze impact factors predictive accuracy. demonstrates precision applicability proposed model.

Язык: Английский

---

DDEYOLOv9: Network for Detecting and Counting Abnormal Fish Behaviors in Complex Water Environments

Fishes, Год журнала: 2024, Номер 9(6), С. 242 - 242

Опубликована: Июнь 20, 2024

Accurately detecting and counting abnormal fish behaviors in aquaculture is essential. Timely detection allows farmers to take swift action protect health prevent economic losses. This paper proposes an enhanced high-precision algorithm based on YOLOv9, named DDEYOLOv9, facilitate the of behavior industrial environments. To address lack publicly available datasets fish, we created “Abnormal Behavior Dataset Takifugu rubripes”, which includes five categories behaviors. The was further several key aspects. Firstly, DRNELAN4 feature extraction module introduced replace original RepNCSPELAN4 module. change improves model’s accuracy for high-density occluded complex water environments while reducing computational cost. Secondly, proposed DCNv4-Dyhead head enhances multi-scale learning capability, effectively recognizes various behaviors, speed. Lastly, issue sample imbalance dataset, propose EMA-SlideLoss, focus hard samples, thereby improving robustness. experimental results demonstrate that DDEYOLOv9 model achieves high Precision, Recall, mean Average Precision (mAP) with values 91.7%, 90.4%, 94.1%, respectively. Compared YOLOv9 model, these metrics are improved by 5.4%, 5.5%, also a running speed 119 frames per second (FPS), 45 FPS faster than YOLOv9. Experimental show can accurately efficiently identify quantify specific

Язык: Английский

---

Estimation of Weight and Body Measurement Model for Pigs Based on Back Point Cloud Data

Animals, Год журнала: 2024, Номер 14(7), С. 1046 - 1046

Опубликована: Март 29, 2024

Pig farming is a crucial sector in global animal husbandry. The weight and body dimension data of pigs reflect their growth development status, serving as vital metrics for assessing progress. Presently, pig dimensions are predominantly measured manually, which poses challenges such difficulties herding, stress responses pigs, the control zoonotic diseases. To address these issues, this study proposes non-contact estimation measurement model based on point cloud from backs. A depth camera was installed above weighbridge to acquire 3D 258 Yorkshire–Landrace crossbred sows. We selected 200 sows research subjects applied filtering denoising techniques three-dimensional data. Subsequently, K-means clustering segmentation algorithm employed extract corresponding pigs’ convolutional neural network with multi-head attention established prediction added RGB information an additional feature. During processing process, we also back size pigs. evaluation, 58 were specifically experimental assessment. Compared manual measurements, exhibited average absolute error 11.552 kg, relative 4.812%, root mean square 11.181 kg. Specifically, MACNN, incorporating feature resulted decrease 2.469 kg RMSE, 0.8% MAPE, 1.032 MAE. Measurements shoulder width, abdominal hip width yielded errors 3.144%, 3.798%, 3.820%. In conclusion, prediction, method demonstrated accuracy reliability measurement.

Язык: Английский

---

Advancing UAV security with artificial intelligence: A comprehensive survey of techniques and future directions

Internet of Things, Год журнала: 2024, Номер 27, С. 101281 - 101281

Опубликована: Июль 6, 2024

Язык: Английский

---