Ensemble and transfer learning of soil inorganic carbon with visible near-infrared spectra

Geoderma, Год журнала: 2025, Номер 456, С. 117257 - 117257

Опубликована: Март 15, 2025

Язык: Английский

---

A novel approach for bearings multiclass fault diagnosis fusing multiscale deep convolution and hybrid attention networks

Measurement Science and Technology, Год журнала: 2024, Номер 35(4), С. 045017 - 045017

Опубликована: Янв. 8, 2024

Abstract Insufficient and imbalanced samples pose a significant challenge in bearing fault diagnosis, leading to low diagnosis accuracy. However, the characteristics of vibration signals are weak difficult extract when faults occur early stage. This paper proposes an effective method that addresses small sample problems under noise interference. First, number faulty form 1D is increased mainly by sliding split sampling method. The preprocessed data used create 2D time–frequency diagrams using continuous wavelet transform (CWT), which can features improve quality. Subsequently, minority oversampled combining synthetic oversampling technique realize conversion augmented oversampling. Moreover, clustering random undersampling introduced prevent overfitting underfitting respectively. Then, we propose hybrid attention mechanism enhance extraction feature information. combination, integrating CWT with multicolumn modified deep residual network, effectively extracts suppresses effects. experimental results demonstrate effectiveness proposed comparison other advanced methods two case studies datasets.

Язык: Английский

---

PRAAD: Pseudo representation adversarial learning for unsupervised anomaly detection

Journal of Information Security and Applications, Год журнала: 2025, Номер 89, С. 103968 - 103968

Опубликована: Янв. 11, 2025

Язык: Английский

---

Attention-Driven Multi-Model Architecture for Unbalanced Network Traffic Intrusion Detection via Extreme Gradient Boosting

Intelligent Systems with Applications, Год журнала: 2025, Номер unknown, С. 200519 - 200519

Опубликована: Апрель 1, 2025

Язык: Английский

---

A Hybrid Slime Mould Meta Heuristic Algorithm and Machine Learning Technique for Intrusion Detection System

Опубликована: Март 14, 2024

Network anomaly prediction is a crucial aspect of network security, as it helps identify unusual or potentially malicious activities within computer network. There are several approaches and techniques used for prediction, Machine Learning (ML) often employed due to its ability detect patterns anomalies in large datasets. In this research, detection carried out by three stages. first stage, data set from IEEE port collected LRC, RFC models trained prediction. second dimension reduction algorithm PCA preprocessing ML third the hybrid +SMA proposed along with From analysis, PCA+SMA+ML provide better performance (accuracy 96) compared previous Finally, PCA+SMA+RFC selected final deployment.

Язык: Английский

---

Enhancing agricultural wireless sensor network security through integrated machine learning approaches

Security and Privacy, Год журнала: 2024, Номер 7(6)

Опубликована: Июль 2, 2024

Abstract Wireless sensor network (WSN) works with a collection of multiple nodes to fetch the data from deployed environment fulfill application whether it is agricultural monitoring, industrial etc. The region can be monitored by deploying verticals where continuous human presence not feasible. These devices are equipped limited resources and easily vulnerable various cyber‐attacks. attacker hack steal critical information WSN devices. cluster heads in play vital role process routing packets attackers launch malicious codes through sender or damage shut down entire regions. This research paper proposes framework improve security WSNs providing shield using machine learning techniques. experimental study includes comparative analysis three techniques decision tree classifier, Gaussian Naïve Bayes, random forest classifier for predicting attacks like flooding, gray hole, blackhole, TDMA that support proposed on attack dataset. achieves an accuracy 98%, Precision 97.6%, Recall F1 score 97.8% which maximum among

Язык: Английский

---

SINNER: A Reward-Sensitive Algorithm for Imbalanced Malware Classification Using Neural Networks with Experience Replay

Information, Год журнала: 2024, Номер 15(8), С. 425 - 425

Опубликована: Июль 23, 2024

Reports produced by popular malware analysis services showed a disparity in samples available for different families. The unequal distribution between such classes can be attributed to several factors, as technological advances and the application domain that seeks infect computer virus. Recent studies have demonstrated effectiveness of deep learning (DL) algorithms when multi-class classification tasks using imbalanced datasets. This achieved updating function correct incorrect predictions performed on minority class are more rewarded or penalized, respectively. procedure logically implemented leveraging reinforcement (DRL) paradigm through proper formulation Markov decision process (MDP). paper proposes SINNER, i.e., DRL-based classifier approaches data imbalance problem at algorithmic level exploiting redesigned reward function, which modifies traditional MDP model used learn this task. Based experimental results, proposed formula appears successful. In addition, SINNER has been compared DL-based models handle skew without relying data-level techniques. Using three out four datasets sourced from existing literature, state-of-the-art performance.

Язык: Английский

---

Convnext-Eesnn: An effective deep learning based malware detection in edge based IIOT

Journal of Intelligent & Fuzzy Systems, Год журнала: 2024, Номер 46(4), С. 10405 - 10421

Опубликована: Март 8, 2024

A rising number of edge devices, like controllers, sensors, and robots, are crucial for Industrial Internet Things (IIoT) networks collecting data communication, storage, processing. The security the IIoT could be compromised by any malicious or unusual behavior on part these devices. They may also make it possible software placed end nodes to enter network perform unauthorized activities. Existing anomaly detection techniques less effective due increasing diversity complexity cyberattacks. In addition, most strategies ineffective devices with limited resources. Therefore, this work presents an deep learning based Malware Detection framework more secure. This multi-stage system begins Deep Convolutional Generative Adversarial Networks (DCGAN) augmentation method overcome issue imbalance. Next, a ConvNeXt-based extracts features from input data. Finally, optimized Enhanced Elman Spike Neural Network (EESNN) is utilized malware recognition classification. Using two distinct datasets— MaleVis Malimg— generalizability suggested model clearly demonstrated. With accuracy 99.24% 99.31% Malimg dataset, strategy demonstrated excellent results surpassed all other existing methods. It illustrates how outperforms alternative models offers numerous benefits.

Язык: Английский

---

Anomaly Detectionin Network Traffic Scenarios by Resampling and Majority Voting with Concept Drift: A Hybrid Approach

2022 10th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO), Год журнала: 2024, Номер unknown, С. 1 - 7

Опубликована: Март 14, 2024

Addressing data imbalance is a critical difficulty in the setting of heavily skewed dataset because dominant class disproportionately influences classifier accuracy, especially when inadequate impede learning minority characteristics. For multi-class cases, traditional binary classification approaches are inadequate. The suggested architecture uses Weighted Majority Voting Classifier (WMVC), noise cleaning, limited under-sampling, and oversampling to overcome this. One important step set an average size limit by dividing entire sample number classes. or more majority classes receive Random Under sampling, all up using Adaptive Synthetic Sampling Approach, reduction achieved via Tomek Link. resulting WMVC contrasted with Convolutional Neural Network (CNN) classifiers, XGBoost, six additional ensemble techniques utilizing tree-based algorithms. Comparative study shows that balanced performs better than unbalanced data. significantly beats CNN, other methods, it improves performance for difficult while successfully decreasing bias towards class.

Язык: Английский

---

Optimal Weighted Voting-Based Collaborated Malware Detection for Zero-Day Malware: A Case Study on VirusTotal and MalwareBazaar

Future Internet, Год журнала: 2024, Номер 16(8), С. 259 - 259

Опубликована: Июль 23, 2024

We propose a detection system incorporating weighted voting mechanism that reflects the vote’s reliability based on accuracy of each detector’s examination, which overcomes problem cooperative detection. Collaborative malware is an effective strategy against zero-day attacks compared to one using only single detector because might pick up overlooked. However, still ineffective if most anti-virus engines lack sufficient intelligence detect malware. Most collaborative methods rely majority voting, prioritizes quantity votes rather than quality those votes. Therefore, our study investigated optimally rates their weight categories expertise engine. implemented prototype with VirusTotal API and evaluated real registered in MalwareBazaar. To evaluate effectiveness detection, we measured recall inspection results same day was MalwareBazaar repository. Through experiments, confirmed proposed can suppress false negatives uniformly improve new types

Язык: Английский

---