This paper examines the use of data flow criteria in software testing and uses evolutionary algorithms to automate the generation of test data with respect to the required k-tuples criterion. The proposed approach is incorporated into an existing test data generation framework consisting of a program analyzer and a test data generator. The former analyses JAVA programs, creates control and data flow graphs, generates paths in relation to data flow dependencies, simulates test cases execution and determines code coverage on the control flow graphs. The test data generator takes advantage of the program analyzer capabilities and generates test cases by utilizing a series of genetic algorithms. The performance of the framework is compared to similar methods and evaluated using both standard and randomly generated JAVA programs. The preliminary results demonstrate the efficacy and efficiency of this approach.