Vattenmärkning av sensordata


  • Karl-Göran Stenborg
  • Jonas Allvar
  • Gustav Haapalahti
  • Magnus Herberthson

Publish date: 2010-12-31

Report number: FOI-R--3138--SE

Pages: 37

Written in: Swedish


  • watermark
  • watermarking
  • fingerprinting
  • digital signature
  • authentication


Sensor data containing audio, image and video are important, and will in future be even more significant, within the Swedish Armed Forces. Therefore, maintenance of confidentiality and integrity of this data material is essential. Watermarking is a method that can be used to embed information in the sensor data. After the embedding, there should be no perceptual difference between the original data and the watermarked sensor data. The watermark can at a later stage be extracted from the sensor data and used for security applications. Within the project "Watermarking of sensor data" three different types of digital watermarks has been considered:  Identification that is used to store information about the sensor data origin or identity. This information will always follow the sensor data, even if it is converted to other formats.  Individual watermarks can be used to trace the source of illicit spread information. Today spreading of secret material is an increasing problem. Such information can for example be spread to websites such as Wikileaks and Openleaks.  Content authentication that is a form of fragile watermarking used to verify the integrity of the sensor data. This can also be used for locating which areas for example an image has been altered. Additionally, you can build in methods to partially restore the damaged areas. This report contains descriptions of the different forms of watermarking. It presents the relationship between the amount of information in the watermark, robustness and distortion that the watermark causes to the sensor data. How individually watermarked sensor data effectively be transferred and stored has been investigated. The demonstration method for content authentication that has been implemented is described. Embedding watermarks are adapted so that the distortion that occurs in the sensor data will be obscure to the human visual sense. Automatic signal processing can be affected by other parameters compared to humans. How a selected detection algorithm is affected if the sensor data has a watermark embedded has been investigated. It turns out that as long as not an unreasonably strong watermark is used, no significant negative impact on the detector from the watermark can be noticed.