Backdoor Detection Using Machine Learning

This research aims to design a backdoor detection technique by using machine learning approach. The implication of this research is to ensure a more effective backdoor detection in network platforms. Security measures against these problems include the use of software programs such as backdoor detection programs. One technique is machine learning (ML); that is a set of tools by which a machine can learn new concepts and new patterns based on a history of learned patterns. The work concentrated on application backdoors which are embedded within the code of a legitimate application. The proposed program in this research is an improvement to backdoor detection based on machine learning techniques. It is developed in Java and employs both supervised and unsupervised methods in the WEKA tool. This helps improving the detection compared to previous methods such fuzzy logic. The results of the experiments have proven that the proposed program is better at detecting backdoors than fuzzy logic when valuated with similar data set. This proves that combining K-Nearest Neighbour and Naive Bayes algorithms is better than using Fuzzy Logic method. The program encountered less false positives and detected all backdoors in the dataset.