The creation of learning materials needs a substantial amount of human expertise. Since the materials are represented as digital data in eLearning systems, the user can make any number of copies without any loss of quality. Therefore, the copyright holders of these materials have a strong interest in protecting their learning objects from illicit use and distribution. One approach to protect intellectual property of such digital contents is digital watermarking. In this paper, a simple wavelet-based watermarking scheme is presented and a comparison of watermark embedding into high and low frequency bands is made. The difference between watermark embedding at first and second level decompositions is also investigated. And finally a framework for optimizing watermark embedding using genetic algorithms is proposed.