Identity document authentication using steganographic techniques: The challenges of noise

Abstract

The term ‘steganography’ encapsulates the practice of secretly embedding data into digital mediums including video, image and audio files. Although steganography is often associated with nefarious activities, conceptually it asserts several characteristics that render it useful in contemporary security applications. Not just a mechanism for criminals to communicate secret information over a digital channel, steganography is also used as a legitimate method of ensuring integrity of digital media artefacts and for identification of same. This application of steganography allows for identification images storing additional information to verify both the identity of the subject as well as the authenticity of the image. Developed methods of steganography invoke various spatial domain techniques that are successful in covertly concealing data within ‘innocent’ carrier images. The techniques include linear methods such as those which replace the least significant bit (LSB) of the bytes in an image and frequency domain methods including discrete cosine transform (DCT), discrete wavelet transform (DWT) and discrete Fourier transform (DFT). The success of a steganographic algorithm is hinged on the method's ability to successfully embed data, so that the data remains concealed within a carrier image; and also to successfully extract the same data uncorrupted. Often modern image coding formats include lossy compression in the frequency domain; this can result in data loss, corruption and noise within the image when carrier images are re-encoded. To ensure data extraction is successful, error correction functions must be invoked to counteract noise and ensure embedded data is extracted without any loss or corruption. In exploring steganographic software, the functionality and reliability of a novel steganographic application ‘Intelligent Identity Authenticator’ (IIA) was assessed. IIA invokes the use of steganography to conceal real-time identity information within images on identity cards. The functionality of IIA is based on a unique algorithm that utilizes DWT to embed a string of characters within an identity image. When data is extracted, it provides a link to further documentation relating to the data subject, allowing for verification of the claimant's identity and authenticity of the identity card. The embedding and extraction functions executed by IIA were found to be mostly reliable, except where data had been embedded within a carrier image that was characterized by a large proportion of black pixels. In these cases, the extracted data string experienced significant loss and corruption, thus preventing access to the identity verification documentation. This paper explores the potential cause of this specific corruption and discusses extensive testing conducted on control images. The results are analysed to identify an improved solution that could rectify the issue, with an aim to improving both the functionality and reliability of the IIA system.