Comparative evaluation of the performance and effectiveness of Machine Learning and Deep Learning algorithms for anomaly detection in Android log analysis
Okangondo Loshima, Serigne Modou Kara Samb, Mohamed Kaba Keita, Doudou Fall, Moussa Dethié Sarr, Idy Diop · 2025
Due to the substantial rise in cyberattacks aimed at Android devices, log analysis has become critical for identifying and mitigating security anomalies. This article conducts a comparative analysis of machine learning and deep learning algorithms for the detection of anomalies within Android logs, specifically sourced from the LogHub dataset. These logs, gathered from Android smartphones equipped with extensive instrumentation, are both rare and intricate. The complexity arises from Android’s multithreaded architecture, making anomaly detection a challenging task.The aim of this study is to identify the most suitable algorithms for detecting anomalies in Android logs, evaluating three algorithms per category (machine learning and deep learning). Using metrics such as precision, recall, F1 score and AUC-ROC, the analysis highlights the effectiveness of each approach depending on context. Finally, recommendations are proposed for optimizing current methodologies and exploring new research perspectives.