© 1979 by British Computer Society
Estimating program reliability
Departments of Electrical Engineering and Computer Science, University of Toronto, Toronto, Canada
This paper describes a process called certification, which can be used to estimate the reliability of a program. Starting from an initial guess at the program reliability, this process uses a controlled series of tests to provide information for a Bayesian updating of the reliability estimate. The test control is derived from a pre/posterior analysis of the two potential decisions of testing a program segment or not and of accepting the program or not, using as criteria the program user's statement of the values to him of the various program outputs. The process has been successfully applied to a few commercial programs in a manual form and is amenable to significant amounts of automation.
Received November 1977.
* Formerly of Dept. of Computer Science, University of Toronto.
Departments of Electrical Engineering and Computer Science, University of Toronto, Toronto, Ontario, Canada