This thesis describes the design and implementation of Byron, an extensible, self-adaptive, general-purpose Evolutionary Fuzzer, capable of fuzzing complex test-program in assembly language. Evolutionary Algorithm (EA) principles and techniques were first described by John H. Holland in the 1960s; in 1990 John Koza described Genetic Programming, a new approach that differs in the type of representation, using trees, and in the application area. From these roots arise various alternative representations of GP, Linear Genetic Programming (LGP), Cartesian Genetic Programming (CGP), and Graph Genetic Programming (GGP); the last one, which is used in Byron implementation. The tool represents the genotypes as Directed Acyclic Graphs (DAGs), which allow the evolution of the graph without going through an intermediate encoding, but directly manipulating the graphs.