Digital image manipulation poses a risk in security and forensics. The existing self-recovery watermarking systems cannot find a balance between the quality of restoration, the localization of tampering, and resistance to extreme geometric attacks. We propose the Dual-Layer Secure Block Mapping (DLSBM) as the solution, a dynamic spatial-domain fragile watermarking architecture. DLSBM incorporates high-fidelity recovery data with a \(2\times 2\) sub-block quantization scheme in \(4\times 4\) blocks, removing classical blocky artifacts. In order to combat the issue of large-area cropping, a new smart center-border mapping plan cross-references the recovery information between each of the vulnerable peripheries and the safe central areas. In addition, adaptive 3-way classification logic distinguishes and recovers from structural alterations, noise, and JPEG compression, utilizing localized median filtering for noise and reserving morphological operations and fast-marching inpainting exclusively for structural artifact removal. Evaluations on standard benchmarks ensure that DLSBM can obtain excellent imperceptibility, precise tamper localization, and high-resolution recovery in various attack conditions. Benchmarked against eleven state-of-the-art fragile self-recovery methods on the USC-SIPI database at exact tampering rates of 10%–50%, DLSBM attains a watermarked-image PSNR of 44.21 dB and a tamper detection rate of 99.98%. Recovered-image PSNR ranges from 35.96 dB at 10% tampering to 26.64 dB at 50% (average 31.79 dB), a graceful 9.3 dB drop across the full range. Overall, the proposed method achieves a robust trade-off between security, accuracy, and visual quality.
