Write a PREreview

This paper presents an applied framework for detecting cyberattacks in embedded systems using machine learning algorithms and signal processing techniques. The study focuses on data collection, preprocessing, and the implementation of artificial intelligence models to identify anomalous behavior in embedded environments. Security-related data were collected through system logs, network monitoring, and sensor readings, followed by the extraction of relevant features using Fourier, Wavelet, PCA (Principal Component Analysis), and Kalman-based signal analysis. The proposed framework integrates both classical and deep learning algorithms, including RF (Random Forest), SVM (Support Vector Machines), K-Means, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), Autoencoders, and GANs (Generative Adversarial Networks), to classify and predict anomalies. The UNSW-NB15 (University of New South Wales – Network-Based Dataset 2015) dataset was used for training and testing, and the obtained results demonstrated high accuracy and robustness against noise and variability. Comparative analysis highlighted the superior performance of ensemble-based methods such as Random Forest and gradient boosting techniques. The experimental findings confirm that the combination of advanced signal processing and AI (Artificial Intelligence)-driven models enables effective identification of malicious network traffic in embedded systems. The proposed framework provides a foundation for real-time, adaptive anomaly detection applicable to resource-constrained environments, supporting the development of resilient embedded cybersecurity.