Emergence of Power-Law Behavior in Defect Distribution of IoT Software

Rapid growth of IoT in various areas is not possible without software. However, IoT software are inherently different from traditional software because of the requirement of supporting multiple types of devices, heterogeneity, distributed and dynamic environment, always-on nature of the system. This has led to altogether new software engineering approach, called IoT-oriented software engineering. Besides, the search for new paradigm of software development, one question worth examining is the distribution of defects in IoT software. The paper investigates the defect distribution in some IoT software to ascertain if they show similar behavior as observed in traditional software or not. To meet this objective, a detailed statistical analysis of defects in IoT software is performed using non-linear regression method. Along with lognormal, Weibull, Pareto and generalized two-parameter Pareto as candidate probability distributions, a new software defect model based on maximum Tsallis entropy framework, called Tsallis distribution, is also included in the study. Tsallis distribution depicts power-law asymptotically. The IoT software are selected from four different categories - in memory database, middleware, data stream processing engine and operating system. Thorough examination of the results of statistical analysis reveal that generalized two-parameter Pareto and Tsallis distributions outperform others leading to the confirmation of emergence of power-law behavior in defect distribution of IoT software a kin to traditional ones. One exception is a relatively new operating system software where lognormal distribution turns out to be best fit. These results augment the IoT-oriented software engineering literature and also help the practitioners in predicting defects in evolving field of IoT software.