Leveraging Fine-Tuned LightGBM for Advanced AI-Driven Android Malware Detection

Increasing instances of cyber threats, specifically malware, require faster creation of proper detection systems that utilize higher learning algorithms. This paper examines the effectiveness of an AI-malware detection technique whereby a LightGBM model finetuned on the CICAndMal2017 dataset is employed. As it will be shown, the CICAndMal2017 dataset is suitable to evaluate the proposed algorithms for detecting malware since this dataset offers sample of different types of malware and benign software. After carefully extracting the features, and selecting the best parameters that allow for the highest potential LightGBM achieved 97.56% accuracy. The results show that the model achieves low false positive rates while accurately identifying the presence or absence of malware in an application. This research helps to advance the cybersecurity field as a whole by presenting a fast, highly efficient, and accurate solution for real-time malware detection with the use of machine learning methods as a proof of concept for eradicating modern cyber threats.