Leveraging Oversampling Techniques in Machine Learning Models for Multi-class Malware Detection in Smart Home Applications - Abstract and Introduction

Aug 5, 2026

Upload by Finn

Smart home applications are increasingly used to deliver safety, comfort, and remote assistance, yet smart home controllers are frequent targets of malware. Successful attacks can cause financial loss, expose personal or sensitive information, and even endanger human lives. While machine learning models have been applied to multi-class malware detection, prior work often overlooks the class-imbalance issue and does not clearly leverage ensemble learners. This study evaluates oversampling methods for minority classes and integrates ensemble models to measure their impact on predictive performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages6

Reading Time15 min

AccessFree to download

Abstract

Introduction

IoT and smart home security risks

Android-based devices as control hubs

Multi-class malware detection and class imbalance

Research contributions

Leveraging Oversampling Techniques in Machine Learning Models for Multi-class Malware Detection in Smart Home Applications Abdullahi Chowdhury School of Computer Science The University of Adelaide Adelaide, Australia [abdul.chowdhury@adelaide.edu.au](abdul.chowdhury@adelaide.edu.au) Mohammad Manzurul Islam dept. of Computer Science and Engineering East West University Bangladesh Dhaka, Bangladesh [mohammad.islam@ewubd.edu](mohammad.islam@ewubd.edu) Shahriar Kaisar dept. of IS and Business Analytics RMIT University Melbourne, Australia [shahriar.kaisar@rmit.edu.au](shahriar.kaisar@rmit.edu.au) Mahbub E Khoda Federation University Australia Churchill, Australia m.khoda,[n.naha@federation.edu.au](n.naha@federation.edu.au) Ranesh Naha Federation University Australia Churchill, Australia [n.naha@federation.edu.au](n.naha@federation.edu.au) Mohammad Ali Khoshkholghi, Mahdi Aiash Dept. of Computer Science Middlesex University London, UKa.khoshkholghi,[m.aiash@mdx.ac.uk](m.aiash@mdx.ac.uk) Abstract—Smarthome applications are becoming increasingly popular due to their ability to provide safety, comfort, and remote assistance. These applications are usually controlled using a smart home controller, which is often the target of malware attacks. A successful attack may result in financial loss, disclosure of personal and/or sensitive information, or even loss of human lives. Although machine learning models have been used in existing research for detecting multi-class malware attacks in smart home systems, they did not explicitly address the class imbalance problem in such cases. In addition, the use of ensemble learner is expected to provide improved performance. To address this, we investigated different oversampling techniques to increase the number of samples in the minority classes and incorporated ensemble learners to see their impact on the prediction performance. Experimental evaluation shows significant improvements (4-5%) in terms of accuracy, precision, recall, and F-1 score. Index Terms—Oversampling Techniques, Ensemble Models, Multi-class Malware Detection I. INTRODUCTION We live in the era of internet-connected smart devices known as the Internet of Things (IoT) . People are connecting their essential everyday devices (e.g., home appliances, wearable sensors, security cameras, intelligent cars) to the Internet for automating intended tasks and making life easy [1] . The scope of IoT device adoption also extends to industrial automation, smart cities, precision agriculture, medical services, etc. A report by Business Insider Intelligence [2] predicts that more than 41 billion IoT devices will be actively connected to the Internet by 2027, resulting in approximately five devices used by each person living on earth. A majority of them are smart home devices that collect, process, and derive insights from sensed data using which smart home sensors decide for the intended action. However, these heterogeneous sensors are generally low-powered with limited processing capabilities and storage, hindering the adoption of a strong security algorithm. Also, to capture the consumer market share, IoT device manufacturers rapidly developed affordable smart home devices without following a standardised security mechanism [3] . Furthermore, for ease of usability by a regular consumer, many connected smart home devices are operated with default authentication, which results in infamous Brickerbot [4], and Mirai [5] attacks. Therefore, an acceptable level of security for IoT smart home devices is far from reality. Among different sensors/devices deployed in smart home applications, android operating system-based devices (e.g., google home, TV, mobile phones) play an important role by acting as a controlling hub, decision processing system, or an end device [6], [7] . However, recent investigations [8], [9] show the vulnerabilities of these android-based devices that can jeopardise the integrity of a smar

FAQs

Why is class imbalance a problem in multi-class malware detection for smart homes?

Training datasets typically contain far fewer malware samples than benign ones, and multi-class labels make learning harder. This imbalance can prevent models from learning minority malware classes effectively.

What does the study investigate to address class imbalance?

Different oversampling techniques are evaluated to increase samples in minority classes. The goal is to find suitable methods that improve performance for multi-class malware detection.

How are ensemble learners used in the proposed approach?

Ensemble learners are incorporated to improve prediction performance. The study assesses the effect of combining oversampling with ensemble models and evaluates metrics such as accuracy, precision, recall, and F-1 score.

Aug 5, 2026

Upload by Finn