Cybersecurity remains a pressing concern in the digital era, necessitating the development of robust defense mechanisms against evolving threats. This study focuses on using advanced machine learning techniques tailored to the dataset’s characteristics for detecting malicious applications in Android devices. The paper presents a novel method of threat detection that identifies features using Principal Component Analysis (PCA) and classifies them using the LightGBM algorithm. The experiment is carried out on network-related properties of adware and benign categories of Android applications obtained from the CICAndMal2017 dataset. The results of the experiments showcase how well this strategy works to identify malicious apps with a high degree of classification accuracy and robust performance. The proposed threat detection framework is promising and very effective in defense of the emerging threats in the Android ecosystem.