The main goal of the thesis is to improve a prototype SBT-Fizzer created in Formela laboratory so that it is able to analyze benchmarks of Test-Comp 2022 competition. There are expected the following key advances in the implementation. The client (analysed program) and server (fuzzer) will be compiled into separate excecutables and they will communicate via the network during the analysis. For each analysed program a new client excecutable will be built before the start of the analysis: the program will be first compiled from C to LLVM bitcode, then instrumented, and finally linked into client's excecutable with an execution driver library. The server must be able to maintain a queue of several spawned client processes in order to amortize costs of starting a single client process and sending data over the network. The improved tool must be evaluated on benchmarks of the latest Test-Comp competition. The results will be presented in the text and tool's strengths and weaknesses must be seen from the data. The functionality of the improved SBT-Fizzer tool and its comparison with other approaches (of tools competing in Test-Comp) will also be described in the text.