A Robust Machine Learning Framework for Fraudulent Mobile App Detection - VFAST Transactions on Software Engineering Volume 12 - Issue 4 - October-December 2024

Aug 4, 2026

Rapid growth of mobile applications has increased the prevalence of fraudulent apps, creating serious exposure to financial loss and potential security compromise. A “Fraud App Detection” framework is proposed to build a reliable system that recognizes and categorizes fraudulent applications using advanced machine learning and artificial intelligence. The workflow gathers data, preprocesses apps, extracts features, trains multiple models, and evaluates performance via recall, accuracy, and F1-score. Neural networks, decision trees, and ensemble methods enhance detection efficiency and accuracy, strengthening mobile app security protocols and safeguarding users from likely threats.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Abstract

Introduction

Problem context: growth of mobile apps and fraud risk

Research goal: AI/ML-based fraudulent app identification

Related approaches: rule-based vs AI/ML methods

Challenges: evolving fraud and class imbalance

Dataset and evaluation preparation

VFAST Transactions on Software Engineering Volume 12, Issue 4, 2024 VFAST Transactions on Software Engineering [https://vfast.org/journals/index.php/VTSE@ 2024](https://vfast.org/journals/index.php/VTSE@ 2024), ISSN(e): 2309-3978, ISSN(p): 2411-6246 Volume 12, Number 4, October-December 2024 pp: 27-36 Keywords: Fraud App Detection, Artiﬁcial Intelligence, Machine Learning, Mobile Applications, Cybersecurity, Fraud Detection, Neural Networks, Decision Trees, Ensemble Methods, Feature Extraction Journal Info: Submitted: October 08, 2024 Accepted: November 08, 2024 Published: November 12, 2024 A Robust Machine Learning Framework for Fraudulent Mobile App Detection Hassan Zaki 1* , Muhammad Saad 1 , Muhammad Rehan Rasheed 2 1 Faculty of Software Engineering, Sir Syed University of Engineering & Technology, Pakistan; 2 Faculty of Computer Engineering, Sir Syed University of Engineering & Technology, Pakistan Abstract The rapid development of mobile applications has led to a signiﬁcant rise in the number of fraudulent applications. The biggest risk now is ﬁnancial loss and possible security compromise. Thus, the "Fraud App Detection" framework goal is to develop a reliable system that can recognize and categorize fraudulent apps utilizing cutting-edge machine learning and artiﬁcial intelligence approaches. The process of identifying fraudulent patterns involves gathering data, preprocessing applications, extracting features, and training several machine learning models. The model’s performance will be assessed based on evaluation criteria like recall, accuracy, and F1-score. To improve detection eﬃciencyand accuracy, this uses cutting-edge techniques such as neural networks, decision trees, and ensemble approaches. These results can be used in enhancing mobile app security protocols, thus safeguarding consumers from the probable threats of fraudulent applications. *Correspondence author email address: [hzaki@ssuet.edu.pk](hzaki@ssuet.edu.pk)[ ](hzaki@ssuet.edu.pk)DOI: 10.21015/vtse.v12i4 .1931 1 Introduction The way we engage with technology has changed dramatically in recent years due to the widespread use of mobile applications (apps), which provide eﬃciencyand convenience in a variety of ﬁelds. Alongside genuine apps, there has been a noticeable increase in fraudulent ones, which pose serious risks like lost money, compromised user privacy, and data breaches. Therefore, it is now critical to identify and remove these fraudulent programmes in order to guarantee the security and reliability of mobile platforms. The goal of this research, "Fraud App Detection," is to use cutting-edge machine learning (ML) and artiﬁcial intelligence (AI) [1–5] techniques to address these issues. The main goal is to create a reliable and precise system that can automatically distinguish and categories phone apps from authentic ones. Through the use of AI and ML, we hope to improve the eﬃcacy andeﬃciency of fraud detection mechanisms in the realm of mobile applications. This exponential growth in mobile applications went hand in hand with increased fraudulent presence within app stores. Concretely, some methods researched by the investigators include sophisticated This work is licensed under a Creative Commons Attribution 3.0 License. VFAST Transactions on Software Engineering Volume 12, Issue 4, 2024 AI and ML algorithms and conventional rule-based systems for fake app identiﬁcation. Early research focused on creating a rule-based system 1 where predeﬁned parameters were implemented that signal unusual behaviors within apps. The new ones utilize AI and ML algorithms, which run through large datasets in search of irregularities that may raise red ﬂags for fraud through reviews by users, app information, and transaction patterns. AI and ML algorithms[6, 7]have shown potential in fraud detection from different areas, such as mobile apps. Some of the used supervised learning algorithms in cate

FAQs

Which datasets are used to construct the training and evaluation set?

Data samples are taken from Kaggle, the UC Machine Learning Repository, and the Canadian Institute for Cybersecurity Dataset IDS 2017, as listed in the document text.

Aug 4, 2026

Upload by Theodore