Table 1 Summary of recent deep learning-based IDS techniques in WSNs (2022–2025).

⁶⁶

Deep Neural Networks (DNN)

Achieved 96.23% accuracy using DNNs on WSN-specific datasets.

High computational cost; limited suitability for real-time systems.

⁶⁷

Deep Learning-based IDS

Demonstrated improved detection performance over classic models.

Risk of overfitting and large data dependency.

⁶⁸

Deep Reinforcement Learning (DRL)

Surveyed DRL-based methods for IoT and WSN intrusion detection.

High implementation complexity; poor scalability in constrained devices.

⁶⁹

Online Ensemble Learning

Achieved 97.2% detection accuracy under dynamic traffic.

May underperform under concept drift.

⁷⁰

LSTM Autoencoder

Outperformed classical ML models on UNSW-NB15 and BoT-IoT datasets.

High resource demand limits use in edge WSNs.

⁷¹

CNN + GANomaly + K-means

Hybrid model boosted detection performance.

Complex architecture; longer training duration.

⁷²

Federated Self-supervised Learning

Privacy-preserving IDS with distributed training.

Increased communication cost and device coordination.

⁷³

ML-based Survey

Comprehensive taxonomy of IDS models for WSNs.

Lacks implementation details and validation.

⁷⁴

Transformer + CNN + VAE-LSTM

Achieved 99.83% accuracy on Kitsune dataset.

Computationally intensive and power-hungry.

⁷⁵

Anomaly Detection Survey

Reviewed anomaly detection techniques for IoT/WSNs.

Broad coverage; limited WSN-specific analysis.