Optimization of Support Vector Machine for Classification of Spyware Using Symbiotic Organism Search for Features Selection

AbstractMalware’s key target is to compromise system security pillars, the confidentiality, integrity and availability. Spyware is a form of malware program that collect entity’s information including personal confidential information, activity logs on computing system, financial transaction, password and geolocation precision through monitoring target without prior knowledge of victim. The integration of computing devices into daily existence, as well as the exponential development experienced in application development including the expansion of interconnected computing devices serve as goldmine to malicious entities for target and exploit using spyware. In previous literature, Support Vector Machine (SVM) was employed for the classification of spyware, but has suffered setbacks of low performance as a result of untuned parameters as well as the use of irrelevant dataset features for training and classification. The optimization of SVM for classification of spyware using Symbiotic Organisms Search (SOS) algorithm for feature selection was therefore deployed to enhance performance. The results obtained from the study indicate that the technique performed optimally in spyware classification recording the following; 97.40% and 2.3% respectively for accuracy and false positive rate respectively. Therefore, revealed that the optimization of SVM with SOS for classification enhances performance and reduces the rate of false alarm which is an improvement on existing literatures. This points the fact that tuned parameters of the model can be implemented for proper classification of spyware attacks.KeywordsSVMSOSSpyware classificationAccuracyFalse positive rate