Visual Security in Mobile Video Applications

Background

Thanks to the technology advances, embedded systems have been widely used in everywhere. In those embedded systems such as smartphones and wearable devices, image and video sensors are utilized to provide various services. Accordingly, many kinds of video applications have been developed. Some of those applications provide entertainments but some other applications deal with privacy-critical and security-critical video data. For example, private video chat between individuals expose their privacy such as private room. Video conferencing has been widely used in business and governmental meeting. In this case, video data could include critical and financial information. In addition, video applications used in military could process extremely classified video data. Video applications used in telemedicine include patient’s privacy and life-critical video information. Therefore, many people concern about privacy and security of important video data in various video applications.

visual_security_fig1

Protect important video data

According to the concerns about privacy and security of important video data, protection techniques to protect those data have been designed. Traditional text encryption algorithms such as DES (Data Encryption Standard) and AES (Advanced Encryption Standard) can be utilized to protect video data. Moreover, various video encryption techniques which consider the characteristics of video data have been proposed. Encryption techniques can be applied before and after the video compression process. Otherwise, techniques could be embedded into the video compression process. Moreover, some techniques protect the entire video data while some other techniques encrypt just several parts of the whole data.

Evaluate visual security of encrypted videos

In case of text encryption, the number of attacks to break the algorithm is used to represent the encryption power of the algorithm in general. However, in case of video encryption, we should consider one more thing, that is visual security of encrypted videos. Visual security represents how much visually distorted when we see the encrypted videos. We can say “the encrypted video A is more visually distorted than the encrypted video B” or “the encrypted video A is heavily distorted while the encrypted video B slightly exposes the frames of the objects”. However, such representations of visual security cannot be used to compare the encryption power of encryption algorithms because each individual have his own criteria for the visual distortion and that is definitely subject, not object. Therefore, we need to represent the visual security of encrypted videos in the objective manner such as numbered levels. PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity) have been widely used to represent visual quality of videos. Moreover, several visual security metrics have been proposed to estimate visual security of encrypted videos.

visual_security_fig2

Energy consumption of video applications with protection techniques

Video coding, which includes video encoding(=compression) and video decoding(=decompression), consumes huge energy due to lots of computations. In addition, video encryption/decryption add more computational overheads, thereby consuming more energy. However, we have limited power resources because mobile devices have limited battery capacity. Moreover, there are several applications and background jobs in mobile devices. Therefore, we should efficiently perform video applications with protection techniques under given energy budget.

visual_security_fig3

Relevant Publications