Title: Automatic Code-Driven Test Generation

Speaker: Patrice Godefroid

Abstract:

Given a program, say with a million lines of C code, wouldn't it be
nice to have a tool that could automatically generate a set of input
values that would exercize, say, even only 50% or 70% of the code in
that program? Today, such a tool does not exist. However, significant
progress towards that goal has recently been made. This new line of
research combines program analysis, testing, model checking and
theorem proving in a novel way. I will describe these latest
developments, with a specific emphasis on current work under way at
Microsoft. For instance, using a recent prototype tool for automatic
code-driven test generation, we have already been able to detect tens
of new expensive security-critical bugs in various Microsoft products.