Table 1.

Comparison of the proposed OGI-based measurement-domain optical steganography with representative GI/SPI-based security schemes in terms of core method, experimental setup, and performance Abbreviations: GI = ghost imaging; CGI = computational ghost imaging; SCGI = sweeping computational ghost imaging; SPI = single-pixel imaging; OGI = orthogonal ghost imaging; FGI = Fourier ghost imaging; CGH = computer-generated hologram; DMD = digital micromirror device; SLM = spatial light modulator; PSNR = peak signal-to-noise ratio; SSIM = structural similarity index measure; NPCR = number of pixels change rate; NRMS = normalized root-mean-square error; CC = correlation coefficient; RSA = Rivest-Shamir-Adleman public-key cryptosystem.

Year/Ref.	Core method	Optical/experimental setup	Security/embedding domain	Representative performance/limitations
202028	Multi-channel computational GI with bucket-sequence permutation and CGI-based steganography (no LSB).	Two-channel computational GI with random patterns and a 3D object; up to $\sim 5.4\times {10}^4$ measurements.	Merged multi-channel bucket vector encrypted via permutation; secret image added to the cover at the object/intensity plane before CGI (not pixel-level LSB).	Provides multi-layer security with good NRMS quality, but requires many measurements and lacks detailed cover-statistics analysis.
202430	Sweeping CGI (SCGI) combined with LSB image steganography and RSA cryptography.	Row/column sweeping SCGI for $64\times 64$ binary/grayscale objects; two bucket sequences from row and column scans.	LSB embedding in SCGI reconstructions (pixel domain); RSA applied to bucket data or reconstructed images.	Few-shot acquisition ($m+n=128$) with good NPCR/NRMS/CC security, but stego-image PSNR remains low (13 dB).
202429	FGI encryption with UTN-designed camouflaged CGHs for holographic steganographic authentication.	DMD-based single-pixel FGI encoder plus dual-SLM optical reconstruction of two cascaded CGHs.	Authentication pattern embedded in CGHs; FGI bucket signals used directly as ciphertext (no bucket-level steganography).	Robust encryption/authentication, but relies on deep-learning CGH design and complex holographic hardware.
202441	Encrypted SPI using Hadamard patterns, 3D Arnold scrambling, and elliptic-curve key exchange.	Hadamard single-pixel imaging; 1D bucket data scrambled by a 3D Arnold transform before transmission.	Pure encryption of SPI light-intensity data (plaintext measurements); no steganographic cover or hidden communication channel.	Enhances encryption security but does not address steganographic invisibility or cover statistics.
202546	Dual-layer optical encryption combining orthogonal ghost imaging with Logistic-map chaos.	OGI with orthogonal patterns; bucket data encrypted digitally using a chaotic keystream.	Bucket-level encryption of OGI intensities before reconstruction; no separate hidden message or steganographic cover.	Provides strong diffusion/confusion but is noise-sensitive and does not address physical-layer steganography.
2025 (this work)	OGI-based Measurement-Domain steganography with hash-initialized logistic-map chaos and bucket-level LSB embedding.	Compact single-pixel OGI with orthogonal patterns and a bucket detector; validated in simulation and optical experiment.	Ciphertext embedded into bucket measurements before OGI reconstruction (hardware-generated OGI cover; no DL or holography).	Maintains PSNR$>53$ dB, SSIM$>0.998$, NPCR$\approx 99.98\%$ while preserving histograms/correlations and lightweight complexity.