A Hybrid Approach to Cloud Database Security: Integrating DL and Machine Learning for Threat Detection and Prevention
Godavari Modalavalasa, Pranay Yadav · 2025
The complexity and ongoing evolution of Advanced Persistent Threats (APTs) compromise the efficacy of conventional cybersecurity measures. Insider threats remain a critical concern in cloud environments, necessitating robust strategies for detection and mitigation. This study presents a new method for detecting intrusions in cloud databases that combines ML and DL approaches. The model used is the Autoencoder Multilayer Perceptron (AE-MLP). This research trains and tests its models using the CICDS 2019 dataset, which includes labeled data on network traffic indicating both typical and malicious activity. The AE-MLP model demonstrates superior performance, achieving 99% accuracy, 98.75% precision, 98.92% recall, and 98.79% F1-score, significantly outperforming other models The findings demonstrate that the AE-MLP model is a viable option for protecting cloud databases since it can identify cyber risks with high accuracy and low false positive and negative rates. The Research article also discusses the performance evaluation, including the training and testing accuracy over epochs and the impact of model complexity on detection efficiency.