Abstract
Geographic Information System (GIS) vector maps have become more widely available, prompting a need to prevent their unauthorized use. This is commonly done through the use of a digital watermark, with many approaches applying techniques from image map watermarking, without exploiting the particular properties of vector map data. In previous work we showed that using k-medoids clustering and the bounding box property of vector maps in the embedding process leads to increased robustness against simplification (removing vertices from vector data) and interpolation (adding new vertices to the data) attacks, which may distort the watermark and prevent the identification of the map owner. In this paper we show that the advantages of using the bounding box property are maintained even with a different clustering approach (k-means), and argue that they would hold regardless of the method used for identifying the watermark embedding locations in the map.