Modeling Software Dependability Growth under Input Partition Testing 
      Yinong Chen                        Jean Arlat
      Dept. of Computer Science          LAAS-CNRS
      Univ. of the Witwatersrand         7 Avenue du Colonel Roche
      Johannesburg, SOUTH AFRICA         31077 Toulouse, FRANCE
      yinong@cs.wits.ac.za               arlat@laas.fr


                                Abstract
One problem of the input domain-based software reliability models is 
the large number of test cases required to obtain a high confidence 
in reliability estimation, because testing has to be restarted from 
beginning after any fault correction is performed. This paper intends 
to overcome this problem by considering relations between the programs 
before and after fault corrections and therefore making use of the 
testing data collected from the previous testing stages. For this 
purpose we propose an input domain-based reliability growth model. 

Both partition and random testing can be used to generate input cases 
for test runs. It is generally considered in the model that input 
generation, fault detection and fault correction are all imperfect. 
It will be shown that some existing reliability models can be viewed 
as the special cases of our model.

Keywords:  reliability modeling, software testing, 
           software dependability, testing coverage