Hiding Functions within Functions: Steganography by Implicit Neural Representations

TL;DR

StegaINR tackles the practical challenge of secret-data extraction without transmitting large decoders by embedding a secret function $f_\theta^m$ into a stego function $f_\phi^c$ constructed via function expansion guided by a shared key $k$. The secret function is recovered by the recipient using $k$ and sampling from the recovered INR, enabling extraction of the secret data $m$ from the stego representation. The method unifies multimedia and 3D data within a single implicit-representation framework, leveraging Random Fourier Features to preserve high-frequency details and achieving high embedding capacity through controlled expansion rate $e = N_{stego}/N_{secret}$. Experimental results across image, climate, and NeRF-like data demonstrate recoverability, varying fidelity tied to expansion, and undetectability against standard steganalysis, highlighting potential for scalable, covert information hiding using continuous-function representations.

Abstract

Deep steganography utilizes the powerful capabilities of deep neural networks to embed and extract messages, but its reliance on an additional message extractor limits its practical use due to the added suspicion it can raise from steganalyzers. To address this problem, we propose StegaINR, which utilizes Implicit Neural Representation (INR) to implement steganography. StegaINR embeds a secret function into a stego function, which serves as both the message extractor and the stego media for secure transmission on a public channel. Recipients need only use a shared key to recover the secret function from the stego function, allowing them to obtain the secret message. Our approach makes use of continuous functions, enabling it to handle various types of messages. To our knowledge, this is the first work to introduce INR into steganography. We performed evaluations on image and climate data to test our method in different deployment contexts.

Figures (14)

• Figure 1: INR-based Steganography: Embedding a Secret Function into a Stego Function. StegaINR enables secure covert communication by embedding and recovering multimodal data via implicit neural representations.
• Figure 2: Steganography Framework based on Implicit Neural Representations.
• Figure 3: Three Strategies for Constructing Stego Functions
• Figure 4: Secret Function Recovery.
• Figure 5: A group of experimental results on the CelebA-HQ database.