Hybrid Control-Flow Checking with On-Line Statistics

One of the methods for improving reliability of embedded systems is control-flow checking (CFC) of the running program. Embedded systems often use hardware or software CFC methods depending on the application. Division of checking parts into software and hardware can increase the system performance when the basic blocks are carefully selected. This paper presents a design of a new hybrid control-flow checking method for detecting errors in control-flow of a running program. The method is designed for programs higher dependent on the inputs. Part of the design is adaptivity that uses on-line statistics of frequently executed basic blocks. Performed experiments show the time overhead with the applied method.