An efficient watermarking framework for multi-face image authentication using chaotic maps and 2D lazy IWT

Abstract The majority of existing face image authentication (FIA) techniques are made for single face image and do not safeguard the multi-face images (MFIs) that are being utilized more and more in biometrics, AI, and surveillance, in addition, the available FIA techniques suffer from limited embedding capacity. Harmonized attacks and manipulations are possible with these MFIs, therefore, this paper draws attention to this research gap and suggests a novel authentication technique. To protect MFIs, a new high-capacity watermarking scheme has been introduced using 2D lazy lifting integer wavelet transform (LL-IWT) and chaotic-based embedding strategy to securely embed authentication and recovery data into non-interest blocks (NIB) of the face image, preserving the integrity of the facial regions. The proposed block-wise segmentation, LL-IWT employment, and the secure embedding strategy guided by chaotic sequences contributed in obtaining significantly high embedding capacity of 1.5 bpp, outperforming prior FIA techniques. Experimental results on diverse datasets of face images demonstrate superior visual quality, achieving an average PSNR of 54 dB and SSIM of 0.997, indicating imperceptibility of the watermark. Moreover, the scheme effectively detects tampering and enables recovery of altered facial regions. These results confirm the scheme’s applicability to real-world scenarios requiring both scalability and security in FIA where it can be used for authenticating single and multiple face images.