Support Vector Machine Classifier Based on Approximate Entropy Metric for Chatbot Text-based Communication

Chatbot is a computer program designed to simulate conversation with human users over the Internet. Chatbot has been found on a number of chat systems, including large commercial chat networks. However, their use as malicious tools has made them a growing nuisance and security concern. We present a support vector machine training algorithm for classification on human and bots in chatbot text-based communications. We use data from the annual Loebner competition to distinguish between bots and humans. The normalized approximate entropy of Message size and inter-message delays at each conversation are introduced. Coupled with the mean and the normalized Shannon entropy of two features, they were considered as the input data. Simulation results have shown that the support vector machine is an efficient method for chatbot data classification.