**9. Detection test against cropping**

One of the geometrical attacks is cropping. In this attack, we manipulated the watermarked video by cropping the video frames. The cropping can be performed horizontally or vertically. In this experiment, we cropped a left side of the video. To extract the

**49**

**10. Discussion**

**Figure 13.**

**Figure 12.**

*right: the tampered region.*

*Application of Chaos-Based Fragile Watermarking to Authenticate Digital Video*

watermarks, we returned the frame size into original size first by adding white or black pixels (in this experiment we added white pixels). We found the extracted watermarks contained the black region that indicated the cropped region in the frames (**Figure 13**).

*The watermarked frames after cropping and the extracted watermarks.*

*Top left: the watermarked frame after changing the brightness; top right: the extracted watermark. Bottom* 

The proposed chaos-based fragile watermarking algorithm is simple but secure; it can be used to authenticate the digital video. Some experiments have been done to

*DOI: http://dx.doi.org/10.5772/intechopen.93151*

*Application of Chaos-Based Fragile Watermarking to Authenticate Digital Video DOI: http://dx.doi.org/10.5772/intechopen.93151*

### **Figure 12.**

*Digital Forensic Science*

video, we got an extracted watermark as shown in **Figure 10b**. The extracted watermark contains a silhouette of strange object inside. Localize the tampered region and we can detect copy-paste object in the video frames as shown in **Figure 10c** and **d**.

There are some kinds of noise such as Gaussian noise, salt and pepper noise, Poisson noise, etc. In the third attack, we added "salt and pepper" noise with density 0.1 into the watermarked video (**Figure 11a**). When we extracted the watermarks from the video, we got the watermarks also contained noise. The noisy watermarks indicated that the video has been altered (**Figure 11b**). The tampering

A digital video can be changed so that the contrast becomes brighter or darker. In this attack, we manipulated the watermarked video by changing the contrast so that make it brighter. After that, we extracted the watermarks from the video (**Figure 12**, top right). We can see that the extracted watermarks are damaged and cannot be recognized anymore. Localization of tampered region shows that whole

*(a) The watermarked frames after adding noise "salt and pepper"; (b) the extracted watermark; (c) and* 

One of the geometrical attacks is cropping. In this attack, we manipulated the watermarked video by cropping the video frames. The cropping can be performed horizontally or vertically. In this experiment, we cropped a left side of the video. To extract the

**7. Detection test against adding some noises**

region is entire of frame (**Figure 11c** and **d**).

**8. Detection test against contrast change**

**9. Detection test against cropping**

of image has been manipulated (**Figure 12**, bottom right).

**48**

**Figure 11.**

*(d) detected tampering region.*

*Top left: the watermarked frame after changing the brightness; top right: the extracted watermark. Bottom right: the tampered region.*

**Figure 13.** *The watermarked frames after cropping and the extracted watermarks.*

watermarks, we returned the frame size into original size first by adding white or black pixels (in this experiment we added white pixels). We found the extracted watermarks contained the black region that indicated the cropped region in the frames (**Figure 13**).

### **10. Discussion**

The proposed chaos-based fragile watermarking algorithm is simple but secure; it can be used to authenticate the digital video. Some experiments have been done to test performance of the algorithm. If there is no manipulation done to the watermarked video (no attack case), then the extracted watermark is same exactly to the original watermark. Therefore, we conclude that the video is still original, has not been changed or manipulated.

Common manipulations of video have been done to test authentication and localize altering in the watermarked video. These manipulations are adding a text label into video frames, inserting a new object into the video, changing contrast, and cropping some pixels. In the case of adding text and inserting an object into the video frames, we got the extracted watermarks that contain silhouette of the object or text. The silhouettes can be seen visually. When we compared to the original watermark, the extracted watermark is not the same. Therefore we conclude that the video has been manipulated. By subtracting the original watermark from the extracted watermark and adjusting the results on the watermarked video, we can find the video frame portion that has been changed.

Common manipulation of video is changing the contrast or brightness of the video. By changing the contrast or brightness of the video, it means changing all pixel values in the video frame. When the watermarks are extracted from the video, we found that extracted watermarks also change entirely. The extracted watermarks are totally damaged; therefore we can conclude that the video has been manipulated.

When a block area in the watermarked video frame is cropped, the extracted watermark is also cropped in the correspondence block. The extracted watermark has a black region in the cropped area of the correspondence frame.

This proposed algorithm has some weakness. It cannot detect manipulation of the watermarked video if one or more fames are removed. However, if some video frames are inserted to the watermarked video, the algorithm can still detect this manipulation, because the new frames do not contain the embedded watermarks.

Other weakness is LSB modification method itself. Bits of the watermark are only embedded to one least significant bit of pixel values. If manipulation of the watermarked video is performed on other than the least significant bit, the algorithm cannot detect it. However, this manipulation is considered uncommon so it can be ignored.
