An Improved Adaptive Watermarking Algorithm for Vector Digital Maps

Han, Pen; Gong, Jianhua; Li, Cheng

doi:10.1109/igarss.2006.731

Cited by 1 publication

(1 citation statement)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some watermarking methods [7,15,16] require the original map or watermarked map as a reference to detect watermarks from a suspicious map. This kind of methods are called "non-blind" algorithms.…”

Section: Non-blind Methods In Spatial Domainmentioning

confidence: 99%

Watermarking road maps against crop and merge attacks

Jiang

Zhu

Huang

et al. 2013

Proceedings of the First ACM Workshop on Information Hiding and Multimedia Security

View full text Add to dashboard Cite

Past research on watermarking digital road maps has been focused on deterring common attacks such as adding noise to the whole map so as to destroy the embedded watermarks. This paper focuses on two less common but increasingly used types of attack: crop attacks and merge attacks. Crop attack crops a fragment of the original map and uses the fragment as a new map. When the new map is much smaller than the original map, it is called massive cropping. Merge attack merges maps from various sources together to form a new map. Conventional watermarking techniques fail against these attacks either because they require global information from the whole map or they must add too many local watermarks and affect the usability of the maps. This paper proposes a novel quad-tree based blind watermarking scheme that partitions the original map according to the quad-tree and plants just one single bit in each sub-region of the map. The approach achieves almost 100% detection accuracy for moderate crop and merge attacks, and over 80% accuracy with more than 95% of the original map cropped and removed. Furthermore, the method introduces very little distortion to the original map: to effectively protect a 23.5MB MinneapolisSt.Paul map against crop and merge as well as other common attacks, only 423 bits or 53 bytes of watermark is required.

show abstract

Section: Non-blind Methods In Spatial Domainmentioning

confidence: 99%