A hybrid CNN-BiLSTM-attention model optimized by an improved sparrow search algorithm for identity authentication and anomaly detection in electricity markets

Conventional rule-based and signature-based approaches struggle to identify sophisticated cyber threats in increasingly digitalized electricity markets, such as identity spoofing, replay attacks, and data tampering, while existing deep learning models often suffer from suboptimal hyperparameter configurations and limited feature representation capability. This work proposes a hybrid deep learning framework that integrates a convolutional neural network, bidirectional long short-term memory, and an attention mechanism for identity authentication in electricity market environments. A bimodal feature engineering scheme is constructed by combining user behavioral characteristics with network traffic statistics, providing complementary information for attack discrimination. To address the sensitivity of deep models to hyperparameter selection, an improved sparrow search algorithm is developed, which incorporates Tent chaotic mapping for population initialization, an adaptive inertia weight for balancing exploration and exploitation, and a Cauchy-Gaussian hybrid perturbation strategy for escaping local optima. Experimental evaluations are conducted on the NSL-KDD and CICIDS2017 benchmarks together with a simulated electricity market dataset built on the IEEE 30-bus test system. On the simulated electricity market dataset, the proposed method achieves an accuracy of 98.71% and an F1 score of 98.43%; averaged across the three datasets it attains 98.31% accuracy, consistently outperforming representative baselines including bidirectional long short-term memory, support vector machine, random forest, XGBoost, convolutional neural network, and long short-term memory. Ablation studies confirm the individual contribution of each component, and robustness analyses show stable performance under the simulated noise levels and attack ratios considered.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-21
DOI
https://doi.org/10.1038/s41598-026-79112-x
Primary Topic
Network Security and Intrusion Detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A hybrid CNN-BiLSTM-attention model optimized by an improved sparrow search algorithm for identity authentication and anomaly detection in electricity markets

Kunhuan Xu, Yang Chaoyi, Yifan Hu, Zhijie Tan et al.
Scientific Reports
Network Security and Intrusion Detection
article

A hybrid CNN-BiLSTM-attention model optimized by an improved sparrow search algorithm for identity authentication and anomaly detection in electricity markets

Kunhuan Xu, Yang Chaoyi, Yifan Hu, Zhijie Tan, Kangqian Huang
article en

Abstract

Conventional rule-based and signature-based approaches struggle to identify sophisticated cyber threats in increasingly digitalized electricity markets, such as identity spoofing, replay attacks, and data tampering, while existing deep learning models often suffer from suboptimal hyperparameter configurations and limited feature representation capability. This work proposes a hybrid deep learning framework that integrates a convolutional neural network, bidirectional long short-term memory, and an attention mechanism for identity authentication in electricity market environments. A bimodal feature engineering scheme is constructed by combining user behavioral characteristics with network traffic statistics, providing complementary information for attack discrimination. To address the sensitivity of deep models to hyperparameter selection, an improved sparrow search algorithm is developed, which incorporates Tent chaotic mapping for population initialization, an adaptive inertia weight for balancing exploration and exploitation, and a Cauchy-Gaussian hybrid perturbation strategy for escaping local optima. Experimental evaluations are conducted on the NSL-KDD and CICIDS2017 benchmarks together with a simulated electricity market dataset built on the IEEE 30-bus test system. On the simulated electricity market dataset, the proposed method achieves an accuracy of 98.71% and an F1 score of 98.43%; averaged across the three datasets it attains 98.31% accuracy, consistently outperforming representative baselines including bidirectional long short-term memory, support vector machine, random forest, XGBoost, convolutional neural network, and long short-term memory. Ablation studies confirm the individual contribution of each component, and robustness analyses show stable performance under the simulated noise levels and attack ratios considered.

Scientific Reports
Ion Exchange (India) (IN), China Southern Power Grid (China) (CN), Power Grid Corporation (India) (IN)
Network Security and Intrusion Detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon