Zero-day exploits detection with adaptive WavePCA-Autoencoder (AWPA) adaptive hybrid exploit detection network (AHEDNet)

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 3, 2025

This paper introduces a new probabilistic composite model for the detection of zero-day exploits targeting capabilities existing anomaly systems in terms accuracy, computational time, and adaptability. To address issues mentioned above, proposed framework consisted three novel elements. The first key innovations are introduction "Adaptive WavePCA-Autoencoder (AWPA)" pre-processing stage which denoising dimensionality reduction, contributes to general dependability accuracy exploit detection. Additionally, "Meta-Attention Transformer Autoencoder (MATA)" enhancing feature extraction subtlety issue, improves model's ability flexibility detect security threats, "Genetic Mongoose-Chameleon Optimization (GMCO)" was introduced effective selection case addressing efficiency challenges. Furthermore, Hybrid Exploit Detection Network (AHEDNet)" dynamic ensemble adaptation issue where is very high with low false positives. experimental results show outperforms other models dataset 1 0.988086 0.990469, precision 0.987976 0.990628, recall 0.988298 0.990435, lowest Hamming Loss 0.011914 0.009531, also, 2 0.9819 0.9919, 0.9868 0.9968, 0.9813 0.9923, 0.0209 0.0109, thus outperformed detecting exploits.

Language: Английский

---

Smart Grid Anomaly Detection Using MFDA and Dilated GRU-based Neural Networks

Smart Grids and Sustainable Energy, Journal Year: 2025, Volume and Issue: 10(1)

Published: Jan. 16, 2025

Language: Английский

---

Adversarial Measurements for Convolutional Neural Network-based Energy Theft Detection Model in Smart Grid

e-Prime - Advances in Electrical Engineering Electronics and Energy, Journal Year: 2025, Volume and Issue: unknown, P. 100909 - 100909

Published: Jan. 1, 2025

Language: Английский

---

An efficient trustworthy cyberattack defence mechanism system for self guided federated learning framework using attention induced deep convolution neural networks

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: May 15, 2025

As cyberattacks become more advanced, conventional centralized threat intelligence models often fail to keep up with these threats' growing complexity and frequency, highlighting the requirement for innovative approaches strengthen cybersecurity resilience. Federated learning (FL), a decentralized machine (ML) model, provides promising solution by permitting spread objects train techniques on local data collaboratively without distributing sensitive data. The efficiency of FL in enhancing attack skills emphasizes its probability driving novel period robust privacy-protecting practices. Furthermore, combining into structures can real upgrades adaptive mechanisms. Recently, ML Deep Learning (DL) have drawn study community advance security solutions cyberattack defence mechanism models. Conventional DL that function kept federal server increase main privacy issues user information. This manuscript presents Cyberattack Defence Mechanism System Framework using Attention Induced Convolution Neural Networks (CDMFL-AIDCNN) technique. CDMFL-AIDCNN model an improved structure incorporating self-guided improve mechanisms across varied applications distributed systems. Initially, preprocessing stage utilizes Z-score normalization transform input beneficial format. Dung Beetle Optimization (DBO) technique is used feature selection process identify most relevant non-redundant features. fusion convolutional neural networks, bidirectional long short-term memory, gated recurrent units, attention (CBLG-A) are employed classify Finally, parameter tuning CBLG-A approach performed growth optimizer (GO) approach. extensively analyzed CIC-IDS-2017 UNSW-NB15 datasets. comparison analysis portrayed superior accuracy value 99.07% 98.64% under

Language: Английский

---

Accurate Power Consumption Predictor and One-Class Electricity Theft Detector for Smart Grid “Change-and-Transmit” Advanced Metering Infrastructure

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(20), P. 9308 - 9308

Published: Oct. 12, 2024

The advanced metering infrastructure (AMI) of the smart grid plays a critical role in energy management and billing by enabling periodic transmission consumers’ power consumption readings. To optimize data collection efficiency, AMI employs “change transmit” (CAT) approach. This approach ensures that readings are only transmitted when there is enough change consumption, thereby reducing traffic. Despite benefits this approach, it faces security challenges where malicious consumers can manipulate their to launch cyberattacks for electricity theft, allowing them illegally reduce bills. While challenge has been addressed supervised learning CAT settings, remains insufficiently unsupervised settings. Moreover, due distortion introduced using accurate prediction future challenge. In paper, we propose two-stage predict detect theft while optimizing For first stage, developed predictor trained exclusively on benign readings, output actual enhance accuracy, cluster-based groups into clusters with similar patterns, dedicated each cluster. second an autoencoder one-class support vector machine (SVM) reconstruction errors classify instances theft. We conducted comprehensive experiments assess effectiveness our proposed experimental results indicate error very small accuracy detection attacks high.

Language: Английский

---

Detection Method for Three-Phase Electricity Theft Based on Multi-Dimensional Feature Extraction

Sensors, Journal Year: 2024, Volume and Issue: 24(18), P. 6057 - 6057

Published: Sept. 19, 2024

The advent of smart grids has facilitated data-driven methods for detecting electricity theft, with a preponderance research efforts focused on user consumption data. multi-dimensional power state data captured by Advanced Metering Infrastructure (AMI) encompasses rich information, the exploration which, in relation to usage behaviors, holds immense potential enhancing efficiency theft detection. In light this, we propose Catch22-Conv-Transformer method, feature extraction-based approach tailored detection anomalous patterns. This methodology leverages both Catch22 set and complementary features extract sequential features, subsequently employing convolutional networks Transformer architecture discern various types behaviors. Our evaluation, utilizing three-phase daily provided State Grid Corporation China, demonstrates efficacy our accurately identifying modalities, including evasion, tampering, manipulation.

Language: Английский

---