A Method for De-Identification Analysis of Encrypted Video

As the variety of video content increases, so does the likelihood of data breaches of sensitive personal information. Consequently, several video de-identification techniques have been developed to protect personal data, alongside ongoing research into video encryption technology, which offers the advantage of decryption to obtain the original video without storing a separate video copy. While it is possible to visually evaluate whether individual encrypted frames have been correctly de-identified, human interpretation makes this method vulnerable to subjective bias. Many video encryption studies employ traditional image quality evaluation metrics, such as peak signal-to-noise ratio and structural similarity index measure (SSIM), to objectively demonstrate the de-identification of encrypted videos. However, these metrics prove to be insufficiently effective for this purpose. These image metrics can measure the de-identification rate of a single frame in encrypted videos, but not the entire video. This study introduces a quantitative analysis method to confirm de-identification in fully encrypted videos. In this study, we measured the de-identification rates of all frames of encrypted videos using a weighted sum of SSIM and edge differential ratio. We also measured the de-identification rate of a fully encrypted video. Finally, we graphically analyzed these data using line charts. This objective methodology enabled the identification of portions of the encrypted video where de-identification techniques may not have been appropriately applied.