An integrated email security gateway and dual-layer frequency-domain watermarking framework for business transaction receipt authentication against generative AI manipulation

Abstract

Business document forgery and email phishing have been significantly amplified by generative artificial intelligence, yet existing security frameworks address communication-channel protection and document authentication as separate concerns, leaving organizations vulnerable to coordinated multi-vector attacks. This research aims to design, implement, and evaluate a unified "Dual-Shield" framework that bridges this gap through an integrated, lightweight solution for real-time business operations. The proposed system combines a four-stage email security gateway—covering sender verification, URL analysis via VirusTotal API, and attachment sanitization—with a dual-layer digital watermarking engine applied to transaction receipt images. The watermarking component pairs a robust Discrete Wavelet Transform–Singular Value Decomposition layer for authorship attribution with a semi-fragile Quantization Index Modulation–Discrete Cosine Transform layer for monetary integrity verification, assessed via imperceptibility, robustness, and tamper-detection metrics on synthetic and real-world datasets. Results show the robust watermark achieved peak signal-to-noise ratio values of 43.38–43.48 dB with SSIM above 0.98. All generative AI manipulations were detected with bit error rates of 0.375–0.472, and targeted ten-percent numerical fraud was identified by the semi-fragile layer. The gateway correctly classified all test scenarios, confirming that malicious URL detection overrides trusted sender status to prevent whitelist exploitation.