|
Original research ANOMALY DETECTION IN INTERNET OF THINGS USING DEEP AUTOENCODERSPages 45-52 Abstract
The rapid growth of IoT networks has revolutionized industries by connecting devices and improving efficiency. However, this expansion has increased vulnerabilities, with traditional anomaly detection methods often struggling to handle the complexity and scale of IoT data. This study addresses these challenges by developing and evaluating a deep autoencoder-based anomaly detection model. The methodology includes data collection, preprocessing (normalization, feature selection), and splitting into training (70%) and testing (30%) datasets. The model leverages an encoder-decoder structure to identify anomalies based on reconstruction errors. Hyperparameter tuning and performance evaluation were conducted using metrics such as accuracy, precision, recall, and F1-score. The model demonstrated strong performance, achieving 90% accuracy, 90% recall, and an F1-score of 0.857. These results highlight its effectiveness in identifying anomalies while maintaining a balance between false positives and negatives. The study provides a robust framework for enhancing IoT network security, addressing real-world challenges, and ensuring reliable, adaptive anomaly detection.
Keywords: IoT networks, Anomaly detection, Deep autoencoder, Reconstruction error, IoT security.
|
Ghuran Marcio 
PDF 