Providing group anonymity in statistical data has become a problem of practical importance. Up to this date, there has been developed a bunch of approaches to anonymizing groups, which are applicable in many real-life applications. It can be shown that sometimes using existing methods is not sufficient to guarantee group data protection, because it is possible to violate anonymity using fuzzy models of given groups in the form of fuzzy inference systems. We propose a memetic approach to providing group anonymity in such cases.