Search Based Software Testing (SBST) and Symbolic Execution (SE) have emerged as the most effective among the fully automated test input generation techniques. However, none of the two techniques satisfactorily solves the problem of generating test cases that exercise specific code elements, as it is required for example in security vulnerability testing. This paper proposes EvoSE, an approach that combines the strengths of SBST and SE. EvoSE implements an evolutionary algorithm that searches the program control flow graph for symbolic paths that traverse the minimum number of unsatisfiable branch conditions. Preliminary evaluation shows that EvoSE outperforms state-of-the-art SE search strategies when targeting specific code elements.