Probabilistic Inference and Trustworthiness Evaluation of Associative Links toward Malicious Attack Detection for Online Recommendations

Abstract

Malware detection systems have faced significant hurdles and strain in recent years due to the fast growth in the quantity and variety of Android malware. Static detection is a popular technique in academics and business for identifying Android malware. However, the current static detecting approaches compromise the unduly high analysis complexity and time cost in order to enhance the detection accuracy. Furthermore, a significant quantity of data becomes redundant due to the connection between static characteristics. As a result, our research suggests a static technique based on sensitive patterns for identifying Android malware. It reduces the creation of superfluous data by mining common combinations of sensitive permissions and API requests in both dangerous and benign applications using an enhanced FP-growth algorithm. Furthermore, the multi-layered gradient boosting decision trees approach is used in this work to train the detection model. Additionally, a dual similarity combination approach is suggested to assess how similar certain sensitive patterns are to one another. The experimental findings demonstrate the excellent generalization capacity and high accuracy of our suggested detection approach.
Key words: ; ;