Modeling A Session-Based Bots' Arrival Process At A Web Server

The paper deals with the problem of modeling key features of the Web traffic generated by Internet bots, observed at the input of a Web server. Based on real log data of an online store, a set of bot sessions was prepared and analyzed. Three session features connected with bots’ arrival process at the server were analyzed: session interarrival time, request interarrival time, and the number of requests in session. Distributional models for these bot session features were developed using regression analysis and validated through graphical comparisons of histograms for the empirical data and simulated values. As a result, interarrival times of bot sessions and interarrival times of requests in bot sessions were modeled by a Weibull and a Pareto distribution, respectively, and the numbers of requests in session were modeled by a function being a combination of a sigmoid and exponential distributions. The aim of our analysis was to develop a model of a session-based bot arrival process on a Web server which may be then implemented in a bot traffic generator integrated with a Web server simulator.