Dynamic Scattering-channel-based Approach for Multiuser Image Encryption

TL;DR

This work addresses the vulnerability of static scattering-based image encryption to learning-based attacks by introducing a dynamic scattering-channel approach for multiuser encryption. It employs time-varying scattering matrices $SM_l$ and user-specific complex coefficients $C(q,q';l)$ to form per-block user keys $K_{q,q'}= \sum_{l=1}^L C(q,q';l)\, SM_l$, with block-level shuffling to enlarge the key space. Decryption reconstructs plaintext via a linear combination of holographic ciphertexts $H^{(l)}$ as $h^{(q\rightarrow q')}=\sum_{l=1}^L C(q,q';l)\, H^{(l)}$, enabling secure multiuser storage and communication through a shared encryption server. Numerical simulations with three users and four scattering states demonstrate that correct keys yield meaningful reconstructions while incorrect or randomly chosen keys yield little to no information, highlighting strong robustness even at modest state counts and indicating scalability as $L$ grows.

Abstract

Conventional scattering-based encryption systems that operate based on a static complex medium which is used by all users are vulnerable to learning-based attacks that exploit ciphertext-plaintext pairs to model and reverse-engineer the scattering medium's response, enabling unauthorized decryption without the physical medium. In this contribution, a new dynamic scattering-channel-based technique for multiuser image encryption is developed. The established approach employs variable, dynamic scattering media which are modeled as tunable aggregates of multiple scattering nanoparticles. The proposed system supports multiple users by allowing distinct combinations of scattering matrices for different time blocks, each combined with user-specific complex-valued coefficients, enabling the creation of unique, hard-to-guess encryption keys for each user. The derived methodology enhances the practical feasibility of multiuser secure communication and storage channels employing scattering media as the encryption mechanism.

Figures (5)

• Figure 1: Schematic of the proposed dynamic scattering-based encryption system.
• Figure 2: (a) Graphical representation of four aggregates of scattering nanoparticles, corresponding to the four distinct states adopted in the simulations, and (b) amplitude of the corresponding scattering matrices.
• Figure 3: Illustration of the dynamic scattering-based encryption scheme for three different users: (a), (b), and (c). The figure shows the images corresponding to reconstruction attempts based on different combinations of the scattered signals $\psi_1$, $\psi_2$, $\psi_3$, and $\psi_4$ corresponding to block scattering matrices $TM_1$, $TM_2$, $TM_3$, and $TM_4$, respectively. Successful reconstruction is achieved only when the correct user-assigned keys are used.
• Figure 4: Analyzing the potential vulnerability of the system to cross-channel interference, where keys from other users are used to reconstruct the original images (a), (b), and (c) of each user. The analysis suggests that no meaningful information about the plaintext images can be extracted using the keys of other users.
• Figure 5: Security analysis of the proposed encryption system. (a) SSIM calculation for twenty random key combinations attempting to hack user 1's information, assuming the intruder has knowledge of the specific scattering matrices used in the hologram generation process. (b) Correct and selected fake keys, represented as $(a, b, c) = (\exp{(-jC_a\pi)}, \exp{(-jC_b\pi)}, \exp{(-jC_c\pi)})$, for cases 2, 5, and 18, are plotted on the complex plane. (c) Visual representation of the reconstructed image for the selected three cases to get a better understanding of the system performance.