• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Friday, November 20, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Technology

Lattice Cryptography Enables Reversible Data Hiding Across Hundreds of Ciphertext Layers

by
November 20, 2026
in Technology
Reading Time: 6 mins read
0
Lattice Cryptography Enables Reversible Data Hiding Across Hundreds of Ciphertext Layers

Lattice Cryptography Enables Reversible Data Hiding Across Hundreds of Ciphertext Layers

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

Computer scientists have developed a post-quantum framework for reversible data hiding in encrypted images that allows multiple independent participants to embed sequential layers of secret messages directly into ciphertexts without altering the carrier or losing the ability to reconstruct the original image. Reporting in the Journal of King Saud University Computer and Information Sciences, researchers Juanli Sun, Minqing Zhang, Yan Ke, Chao Jiang, Lin Shi, Mingshu Zhang, and Siyuan Huang demonstrated that structured lattice cryptography based on the Ring Learning with Errors problem can support up to 255 sequential embedding layers. Conventional reversible data hiding methods operate either through single-embedding procedures or restrict multi-layer embedding to two or three phases, often relying on cryptosystems that remain vulnerable to quantum cryptanalysis. By leveraging the additive homomorphic properties of cyclotomic polynomial rings and establishing an adaptable coefficient-partitioning mechanism, the investigators demonstrated a scalable architecture suited for decentralized, multi-party processing environments such as collaborative medical consultations and distributed data tracking.

The engineering challenge addressed by the team centers on maintaining complete reversibility when numerous independent entities must annotate shared, confidential records. Furthermore, while secret sharing and public-key encryption schemes like Paillier have been adapted for multi-user tasks, they either require distributing data shares among all actors simultaneously or introduce severe computational bottlenecks and structural susceptibility to quantum Shor attacks. Earlier post-quantum attempts employing standard Learning with Errors mechanisms enabled limited multi-stage processing, yet they restricted embedding capacity and critically assumed that intermediate data hiders possessed decryption keys. In contrast, the newly formulated scheme strictly segregates administrative capabilities, ensuring that intermediate data hiders require only the public cryptographic parameters and individual embedding keys to insert annotations while preserving overall confidentiality.

To establish this capability, the authors structured their cryptosystem over a cyclotomic polynomial ring denoted as the quotient ring of integers modulo q divided by x to the power of n plus one, setting the security parameter ring dimension to n equals 256 and selecting an odd modulus q equal to 57857. In this configuration, plaintext vectors are treated as polynomials in binary coefficients, while ciphertexts comprise pairs of ring elements modulo q. The design partitions the n available polynomial coefficients into l plus one disjoint, non-overlapping intervals, allocating an initial segment of length n0 for plaintext image block storage and subsequent segments of length ni for each respective embedding tier up to the total layer count l. When generating keys, the public parameters define discrete Gaussian error distributions with a standard deviation derived from a noise parameter alpha approximately equal to 4.8887 times ten to the negative fifth power, satisfying the asymptotic hardness condition necessary to preserve the worst-case to average-case lattice reduction guarantees of Ring Learning with Errors.

The workflow begins at the content owner, who decomposes an uncompressed grayscale image into its constitutive eight bit-planes and partitions the raster stream into blocks matching the allocated plaintext coefficient length n0. Each image block vector is zero-padded up to the full dimension n by appending trailing zeros, transforming the block into a polynomial mx. The content owner encrypts each polynomial by drawing random noise polynomials from the discrete Gaussian distribution and generating a ciphertext pair using the public key. This encrypted carrier can then be transmitted across open networks to successive data hiders. Each intermediate participant converts their private binary payload into segments of length ni, aligns the segment with their designated coefficient interval through appropriate leading and trailing zero-padding, and blinds the payload via bitwise XOR with a pseudorandom sequence derived from an individual timestamp embedding key. The blinded payload polynomial is independently encrypted under Ring Learning with Errors and added directly to the existing image ciphertext through homomorphic polynomial addition.

Because homomorphic additions linearly compound inherent noise polynomials across every consecutive embedding stage, the authors conducted an inductive derivation to rigorously bound the cumulative noise term. For each embedding stage, the newly introduced ciphertext contributes fresh, independent Gaussian error terms originating from newly sampled polynomials. Applying the Lindeberg-Levy Central Limit Theorem across progressive embedding cycles, the researchers established that each coefficient of the aggregated noise term asymptotically converges to a zero-mean Gaussian distribution with a per-layer coefficient variance determined by the system modulus, ring dimension, and Gaussian noise parameter. Decryption correctness requires that the infinity norm of this cumulative noise polynomial remain strictly below the decision boundary of floor q divided by 4, which equates to 14464 for a modulus of 57857. Through a deterministic Gaussian tail bound, the team verified that the overall theoretical failure probability across all n coefficients remains below 1.57 times ten to the negative fourth power for the maximum load of 255 layers, while empirical testing across 1000 benchmark images observed zero decryption failures.

To evaluate the practical capabilities and limits of the protocol, the team conducted extensive trials on 1000 grayscale images sized at 512 by 512 pixels taken from four standardized repositories: USC-SIPI, BOSSBase, and the medical imaging databases OASIS and Lung-PET-CT-Dx. Executed on a test workstation equipped with an Intel Core i7-8565U processor and 8 gigabytes of RAM running Python 3.10, the system demonstrated complete message extraction and flawless carrier reconstruction. Across configurations spanning 1 layer up to 255 layers on standard test images such as Lake and Man, the extraction accuracy for both the initial layer and the terminal layer consistently reached 100 percent. Following modular thresholding against zero and floor q divided by 2, bitmasking and shifting operations successfully separated the decrypted carrier stream from the blinded payloads, enabling the receiver to invert the pseudorandom sequences and reconstruct the original carrier image with peak signal-to-noise ratio values approaching infinity.

The quantitative evaluation revealed that the plaintext embedding rate scales linearly with the number of embedding tiers, governed by the ratio of message coefficients to plaintext coefficients. Operating under an equal partition allocation where n0 equals floor n divided by l plus one, the plaintext embedding rate begins at approximately 8 bits per pixel for a single layer and reaches 2040 bits per pixel at 255 layers, achieving the theoretical ceiling where only a single coefficient per polynomial is reserved for the initial image. In comparative benchmarks against existing reversible data hiding implementations, the scheme surpassed conventional stream-cipher approaches, which achieved embedding rates between 4.78 and 4.81 bits per pixel, as well as secret-sharing methods yielding between 0.90 and 3.64 bits per pixel. Paillier-based methods reached between 13.48 and 1029 bits per pixel but remained vulnerable to quantum cryptanalysis, while earlier post-quantum Learning with Errors systems reached a ceiling of 24.50 bits per pixel across only two or three supported layers.

The experimental analysis also characterized the trade-offs imposed by ciphertext expansion and computational runtimes across different system parameters. Because each one-bit plaintext coefficient is transformed into a pair of polynomial elements represented modulo q, the ciphertext expansion factor rho increases with the number of layers, scaling from 64 at layer 1 to 8192 at layer 255. Consequently, the ciphertext storage for a 256-kilobyte grayscale image ranges from 16 megabytes up to 2 gigabytes, resulting in an effective ciphertext embedding rate between 0.0156 and 0.0311 bits per ciphertext bit. Despite this expansion, the computational performance of the ring structure proved significantly faster than modular exponentiation in Paillier schemes and matrix multiplications in standard Learning with Errors. Under a fixed two-layer configuration, the total execution time across encryption, embedding, decryption, and extraction was 138.25 seconds for the proposed method, compared to 1748.06 to 1992.53 seconds for Paillier-based approaches and 358.17 to 865.83 seconds for conventional Learning with Errors designs.

Formal security proofs confirmed that the proposed framework achieves semantic indistinguishability under the Decisional Ring Learning with Errors hardness assumption, meaning an adversary cannot distinguish marked ciphertexts from uniformly random ring elements. Statistical validations demonstrated that encrypted images exhibit uniform distributions, with information entropy values approaching the theoretical ceiling of log2 q at approximately 15.8202 bits, remaining indistinguishable before and after multi-layer embedding. Furthermore, Pearson correlation coefficients measured across horizontal, vertical, and diagonal directions dropped from high values near unity in plaintext images to near zero in ciphertexts. Additional robustness tests revealed that the system tolerates substantial external additive Gaussian noise, maintaining zero bit error rates and 100 percent extraction accuracy up to an added noise standard deviation of approximately 1000 before gradually degrading as perturbations approach the theoretical boundary.

While the architecture successfully resolves the layer and capacity bottlenecks of multi-party encrypted data hiding, the researchers noted that practical real-world deployments must carefully calibrate the layer depth against bandwidth and transmission constraints. In network-constrained settings, setting a moderate parameter such as l equals 7 restricts the ciphertext size to 64 megabytes while still delivering an embedding capacity of 56 bits per pixel. The team highlighted that the current primitives operate sequentially without enforcing cryptographic access controls or preventing unauthorized intermediate parties from altering embedding order. To build upon these findings, the authors plan to investigate multi-key fully homomorphic encryption to support decentralized keys among multiple owners, alongside fine-grained access control mechanisms that enforce hierarchical embedding privileges and differentiated data visibility across distributed environments.

Subject of Research: Post-quantum reversible data hiding in encrypted images using Ring Learning with Errors

Article Title: Reversible data hiding in encrypted images with multi-layer embedding based on ring learning with errors

Article References: Sun, J., Zhang, M., Ke, Y., Jiang, C., Shi, L., Zhang, M., & Huang, S. (2026). Reversible data hiding in encrypted images with multi-layer embedding based on ring learning with errors. Journal of King Saud University Computer and Information Sciences, 38(9), Article 1013. https://doi.org/10.1007/s44443-026-01387-x

Image Credits: AI Generated

DOI: 10.1007/s44443-026-01387-x

Keywords: Reversible Data Hiding, Ring Learning with Errors, Ciphertext Expansion, Post-Quantum Cryptography, Additive Homomorphism, Encrypted Images, Multiparty Communication, Lattice-Based Cryptography, Information Entropy, Bit-Plane Decomposition, Decryption Correctness, Medical Consultation Security, Polynomial Ring Encryption, Cumulative Noise Bound, Lossless Image Recovery, Data Privacy Protection

News Source: Denise Maddox. (November 20, 2026). Lattice Cryptography Enables Reversible Data Hiding Across Hundreds of Ciphertext Layers. Scienmag.

Tags: Additive HomomorphismBit-Plane DecompositionCiphertext ExpansionCumulative Noise BoundData Privacy ProtectionDecryption Correctnessencrypted imagesinformation entropylattice-based cryptographylossless image recoveryMedical Consultation SecurityMultiparty CommunicationPolynomial Ring Encryptionpost-quantum cryptographyreversible data hidingRing Learning with Errors
Share12Tweet7Share2ShareShareShare1

Related Posts

Displacement Functional Minimization Establishes Meir–Keeler Fixed Points Without Space Completeness

Displacement Functional Minimization Establishes Meir–Keeler Fixed Points Without Space Completeness

November 20, 2026
Borehole Constraints Eliminate Deep Ground Velocity Artifacts Caused by Seasonal Freezing

Borehole Constraints Eliminate Deep Ground Velocity Artifacts Caused by Seasonal Freezing

November 20, 2026

Tunnel Model Tests Reveal Structural Limits of Prefabricated Segment Lining Schemes

November 20, 2026

Review Assesses Photonic Crystal Logic Gates Across Lattice Geometries and Material Platforms

November 20, 2026

POPULAR NEWS

  • Pediatric Urology AI Prediction Models Lack External Validation Despite High Discrimination

    Pediatric Urology AI Prediction Models Lack External Validation Despite High Discrimination

    29 shares
    Share 12 Tweet 7
  • Ultraviolet-Ozone Fabric Treatment Boosts 3D-Printed Polymer Adhesion by Up to 239 Percent

    29 shares
    Share 12 Tweet 7
  • Rapid Review Identifies Thirty-Eight Parent-of-Origin Genetic Variants Linked to Fetal Growth

    29 shares
    Share 12 Tweet 7
  • Breeding Failure Prompts Earlier Autumn Migration in Southern British Eurasian Curlews

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Augmented Reality Navigation Shows Feasibility in Pilot Brain Biopsy Study

Displacement Functional Minimization Establishes Meir–Keeler Fixed Points Without Space Completeness

Bronchiectasis Affects Over a Quarter of Patients With Inborn Immune Errors

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 89 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.