|
AUTHOR(S): Bartholomew Idoko, Francisca Ogwueleka, Steven Bassey, Monday Adenomon
|
TITLE |
 PDF |
ABSTRACT The cyberspace is presently faced with many incidences related to malware compromising the cyber security goals of; confidentiality, integrity, availability, authenticity, non-repudiation and trust of system networks. Malware authors are breaking frontiers in developing new kind of malware such that anti-virus software cannot provide the level of protection that is being anticipated by users thereby resulting in unprecedented successful attacks being recorded. The study examine how existing machine learning models can be utilized in classifying malware infections and its propagation mechanism. This study adopts an experimental and quantitative research design using Machine Learning techniques to develop and evaluate a hybrid malware detection model. The approach includes the data collection, feature engineering, model development and performance evaluation. The hybrid model is designed using the combination of both malware (infected) files and benign (uninfected) files. The implementation of the system is done with the aid of deep learning structure (Inception v3) with SVM classifier by making use of the CNN learning method. The framework’s input comes from instances of the explored malware dataset. These instances are supplied through the hybrid model (Inception v3) input layer. The instances are further forwarded to convolutional phase of Inception v3, where important features are extracted. The extracted features are utilized in form of support vector machine classifier’s input. The incorporated support vector machine classifier uses the extracted features for classification where the output of the hybrid system is determined as either malware file or legitimate file. The hybrid model is made up of two different components; Inception v3 and SVM which combines to produce the hybrid model using the malware dataset to constructively detect malware. The final module, SVM, is used for classification, whereas Inception v3 is used for feature extraction. The simulation of all the tested models is carried out in order to analyze the data and answer the research questions. These models include; the developed hybrid model (Inception v3 + SVM), Convolutional Neural Network, CNN, Support Vector Machine, SVM, and Neural Network, NN. The aim is to deduce the best model for the malware dataset as well as to determine the best performing model. The accuracy rate of the hybrid system was obtained as 99.89%. The result shows significant convergence in both performance and learning. The result undoubtedly show how effective the proposed hybrid design is in comparison to other models developed to address the same issue. |
KEYWORDS Malware, Detection, Model, Machine learning, Accuracy, Classification, Hybrid |
|
Cite this paper Bartholomew Idoko, Francisca Ogwueleka, Steven Bassey, Monday Adenomon. (2025) A Hybrid Machine Learning Model for Malware Detection. International Journal of Internet of Things and Web Services, 10, 44-62 |
|
