Abstract: A national identity document is an identity card with a photo, usable as an identity card at least inside the country, and which is issued by an official authority. The most common applications for these smart cards are smart to travel documents, electronic IDs, electronic signatures, municipal cards, key cards used to access secure areas or business infrastructures, social security cards, etc. These documents have several security features which mitigate and combat document forgery. As these security systems are difficult to circumvent, criminal attacks on ID verification systems are now focusing on fraudulently obtaining genuine documents and the manipulation of the facial portraits. Trusted identity is a vital component of a well functioning society. To reduce risks related to this fraud problem, it is necessary those governments and manufacturer of IDs continuously develop and improve security measures. With this in mind, we introduce the first efficient steganography method – StegoCard – which is optimized for facial images printed in common IDs. StegoCard is an end-to-end facial image steganography model that is formed by n Deep Convolutional Auto Encoder, that can conceal a secret message in a face portrait and, hence, producing the stego facial image, and a Deep Convolutional Auto Decoder, which is able to read a message from the stego facial image, even if it is previously printed and then captured by a digital camera. Facial images encoded with our StegoCard approach outperform the StegaStamp generated images in terms of their perception quality. Peak Signal-to-Noise Ratio, hiding capacity and imperceptibility results on the test set are used to measure the performance.
|
DOI:
10.17148/IJARCCE.2023.12553
