Year: 2024
Type: article
Abstract: Nowadays, everybody desires to possess their personal mobile device, and this has led to an increase in the number of Android users worldwide. Every device connected to the internet communicates with a multitude of applications and thus ends up having many malware attacks or threats in a mobile home. This paper takes a comparative design of predictive models based on smartphone detecting communication anomalies through AndroVul. In this paper, a framework is proposed to be developed to detect cases of communication abnormality using the AndroVul data set with the help of an Artificial Neural Network (ANN). The methodology consists of the preprocessing, feature selection and model training to do a good differentiation of the benign and malicious communication behavior. The evaluation conducted on the experiment indicates that ANN performs with 99.33% accuracy (ACC) , 99.71% precision (PRE), 99.65% recall (REC) and an F1-score (F1) of 99 which is superior to the classical models: MLP (84.31%), Random Forest (93.6%), PCA-based classifiers (89.3%), and SVM (98.2%). Such results indicate the efficiency of AI-based solutions to increase the security of communication in smartphones, and can be applied on a large scale to identify potential abnormalities of communication in the mobile environment. In general, the most important contribution is the introduction of a strong AI-based framework that improves the security of mobile communication by providing reliable, scalable, and high-performance anomaly detection.
Institution: Microsoft (United States)
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7142121111