A Novel Software Reliability Growth Model Based on Generalized Imperfect Debugging NHPP Framework

Abstract

Non-Homogeneous Poisson Process (NHPP) is a standard framework in the field of software reliability analysis. The core of NHPP consists in determining the Mean Value Function (MVF) of cumulative error number at a specific time slot. However, practice shows the difficulty in finding a general model to fit all sorts of fault data. A certain model is only sensitive to the specific object(s). Modeling failure MVF for NHPP still faces a number of challenges such as making reasonable explanation of assumption, determining fault detection rate per error, fault modification efficiency, error introduction rate, etc. In this research, we propose a novel Software Reliability Growth Model (SRGM) by leveraging generalized imperfect debugging NHPP framework. We first provide physical explanations for assumptions on error modification, error introduction and fault detection rate per error. Meanwhile, we generate a typical constraint relationship between the total error introduction rate and change rate of generalized residual errors. We also describe the fault detection rate per error with the form of exponential decay function, and use error reduction factor to form the new model. Furthermore, we make extensive discussions based on our proposed model. The experimental results confirm that our proposed model is effective on fault fitting and prediction, especially excellent on short-term prediction.

Keywords

• Software reliability growth
• non-homogeneous Poisson process
• error introduction rate
• fault detection rate per error
• error reduction factor
• mean value function