Challenges in Software Certification

As software has invaded more and more areas of everyday life, software certification has emerged as a very important issue for governments, industry and consumers. Existing certification regimes are generally focused on the wrong entity, the development process that produces the artifact to be certified. At best, such an approach can produce only circumstantial evidence for the suitability of the software. For proper scientific evaluation of an artifact, we need to address directly the attributes of the product and their acceptability for certification. However, the product itself is clearly not enough, as we need other artifacts, like requirements specifications, designs, test documentation, correctness proofs, etc. We can organise these artifacts using a simple, idealised process, in terms of which a manufacturer’s own process can be “faked”. The attributes of this idealised process and its products can be modelled, following the principles of Measurement Theory, using the product/process modelling method first introduced by Kaposi.