CoChat
Sign Up
Cited by 0Chaos-based Image/Signal EncryptionCryptographic Implementations and SecurityComputer Science and Engineering

Development of image encryption method using surjective finite automata and custom S-box within the advanced encryption standard framework

Open access
Barlybayev, Alibek, Saukhanova, Zhanat, Shakhmetova, Gulmira, Sharipbay, Altynbek, Raykul, Sayat, Khassenov, Altay·2025·Eastern-European Journal of Enterprise Technologies
DOI 10.15587/1729-4061.2025.348368
Open this paper in CoChat — free accountOpen PDF

Abstract

The object of the study is the AES-128 (Advanced Encryption Standard) – based image-encryption scheme. The problem solved is the persistence of residual image structure and sub-ideal statistical security when classical AES is naively applied to visual data. The generated cipher images yield near-maximal entropy with low pixel correlations and uniform histograms. Encrypted-image Chi-square values concentrate around 200–310 (close to a uniform distribution), the NPCR (Number Of Changing Pixel Rate) consistently 99.623–99.657% with a best case 99.6547%, and the UACI (Unified Averaged Changed Intensity) ≈ 33.64% per channel (RGB combined ≈ 22%). Robustness tests show ≈ 30.7 dB at 50% cropping and ≈ 39.5–39.7 dB at 0.01 salt-and-pepper noise and 6.25% cropping. These outcomes are explained by the bit-level, state-dependent permutations introduced by the surjective automaton (boosting diffusion) and by the nonlinear S-box synthesized under strict criteria (e. g., bounded differential uniformity, high nonlinearity) that heighten confusion, and operation in CBC (Cipher Block Chaining) mode supplies semantic security. Other unique features that facilitate the solution are the substituting of ShiftRows/MixColumns with surjective finite automata; a custom, criteria-optimized S-box; and a 10-round AES-128 CBC pipeline with a random. Taking all of this together yielding observed statistical uniformity, a high NPCR/UACI, and stable robustness under degradation. Lastly, the findings demonstrate the applicability to secure multimedia transmission and storage in channels prone to noise or partial data loss, and being data-agnostic, that the transformations can generalize to text and generic binary data when carefully managed

AI snapshot

Written from the paper’s full text.

TL;DR

The paper proposes an AES-128 image-encryption variant that replaces ShiftRows/MixColumns with a surjective finite automaton and the Rijndael S-box with a custom criteria-optimized 8-bit S-box in a 10-round CBC pipeline. It reports near-uniform histograms, entropy ≥7.99, NPCR around 99.623–99.657%, UACI around 33.64% per channel, and stable PSNR under cropping and salt-and-pepper noise.

Key findings

  • Encrypted images showed entropy above 7.99 bits per channel, chi-square values concentrated around 200–310 (e.g., “4.1.05Blue” = 300.57), and correlations near zero, with Mandrill red changing from r ≈ 0.9217 to 0.0034, green from 0.8643 to −0.0015, and blue from 0.9071 to −0.0006.
  • NPCR ranged from 99.623% to 99.657%, with a best case of 99.6547%; UACI was ≈33.64% per channel (RGB combined ≈22%), compared with DES NPCR 99.5911%/UACI 15.0325% and RC4 NPCR 99.5895%/UACI 14.8831%.
  • Robustness tests gave PSNR ≈30.7 dB at 50% cropping, ≈39.5–39.7 dB at 0.01 salt-and-pepper noise and 6.25% cropping, 33.1 dB at 0.05 noise, and 29.2 dB at 0.2 noise.
  • All fifteen NIST STS categories passed at α = 0.01 with p-values > 0.01 for the binary expansion of Euler’s number (e), as reported in the paper.

Methods

  • Design: modified AES-128 with 10 rounds in CBC mode, 128-bit blocks and keys; ShiftRows/MixColumns replaced by a surjective finite automaton A = ⟨Q, X, Y, δ, λ⟩; SubBytes replaced by a custom 8-bit S-box generated via a generalized algebraic method plus gradient descent, with criteria including nonlinearity ≥100, algebraic degree ≥7, differential uniformity ≤8, and no fixed or opposite fixed points.
  • Data: eight USC-SIPI images — four 256×256 images (“4.1.04 Female”, “4.1.05 House”, “4.1.06 Tree”, “4.1.08 Jelly Beans”) and four 512×512 images (“4.2.03 Mandrill”, “4.2.05 Airplane”, “4.2.06 Sailboat on Lake”, “4.2.07 Peppers”) — processed as 8-bit RGB TIFF inputs in 128-bit blocks.
  • Analysis: 2D/3D correlation coefficients, RGB-split and merged histograms, entropy, chi-square goodness-of-fit, arithmetic mean, NIST SP 800-22, NPCR/UACI, and PSNR under block-wise occlusion and salt-and-pepper noise; comparisons were made with AES, DES, RC4, RSA, and other published methods.

Limitations

  • The paper states that the application was demonstrated only on RGB images from USC-SIPI; generalization to medical grayscale, HDR, hyperspectral, and compressed formats such as JPEG was not tested.
  • The paper states that reproducibility depends on random NP exchanges during S-box synthesis and random FA state construction, so repeated generation may yield different instances with slightly different statistics.
  • The paper states that security analysis is based on statistical metrics and PSNR, with no formal cryptanalysis of linear, differential, related-key, boomerang, or chosen-ciphertext attacks; it also notes increased implementation complexity, possible overhead versus hardware-accelerated AES, and the need for constant-time side-channel protection.
  • Evident from the text: the NIST STS results are reported for the binary expansion of Euler’s number (e) rather than clearly for the cipher images, and no throughput/latency benchmarks are presented despite the paper’s discussion of performance parity gaps.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.15587/1729-4061.2025.348368

Cites

42 of 42 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 13 cited sources checked · 0 supported · 0 flagged

  • Findingreasoning

    Robustness tests show approximately 39.5-39.7 dB at 0.01 salt-and-pepper noise and 6.25% cropping.

    “≈ 39.5-39.7 dB at 0.01 salt-and-pepper noise and 6.25% cropping”
    Quote in the paperSubstantial claim
  • Findingreasoning

    Encrypted-image Chi-square values concentrate around 200-310, close to a uniform distribution.

    “Encrypted-image Chi-square values concentrate around 200-310 (close to a uniform distribution)”
    Quote in the paperSubstantial claim
  • Findingreasoning

    Robustness tests show approximately 30.7 dB at 50% cropping.

    “Robustness tests show ≈ 30.7 dB at 50% cropping”
    Quote in the paperSubstantial claim
  • Findingown data

    The PSNR is approximately 30.7 dB at 50% cropping.

    “The PSNR is approximately 30.7 dB at 50% cropping.”
    Quote in the paperSubstantial claim
  • Findingown data

    Textured images such as Mandrill and Peppers show large drops in correlation after encryption.

    “Textured cases such as Mandrill and Peppers show large drops in correlation.”
    Quote in the paperSubstantial claim
  • May combine two claims. 33
  • Classified as probably not a substantial claim. 23
  • The quote only partly states this claim. 2

  • May combine two claims.

    Findingreasoning

    The generated cipher images yield near-maximal entropy with low pixel correlations and uniform histograms.

    “The generated cipher images yield near-maximal entropy with low pixel correlations and uniform histograms.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    NPCR is consistently 99.623-99.657%, with a best case of 99.6547%.

    “the NPCR (Number Of Changing PixelRate) consistently 99.623-99.657%with a best case 99.6547%”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    UACI is approximately 33.64% per channel, with RGB combined approximately 22%.

    “the UACI (Unified Averaged Changed Intensity) ≈ 33.64% per channel (RGB combined ≈ 22%)”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    At noise density 0.01, decrypted images still keep quality; at 0.2, noise takes over and image becomes hard to read.

    “At a noise density of 0.01, the decrypted images still keep the quality.”
    • Also stated in: § Continuation of Table 7 · plain
      “At a density of 0.2, the noise takes over.The image becomes hard to read.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In contrast to [43], swarm-based methods have UACI that only reaches 30.147%, and this result has UACI close to the optimum of about 33%.

    “In contrast to [43], swarm-based methods have UACI that only reaches 30.147%, and this result has UACI that's close to the optimum of about 33%.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [43]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Findingown data

    The proposed method shows an NPCR value of 99.6547% and a UACI value of 33.6374%.

    “The proposed method shows an NPCR value of 99.6547% and a UACI value of 33.6374%.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In contrast to DES, RC4 and SIT [42], where UACI values are 14-15%, this UACI is 33.637433%.

    “In contrast to DES, RC4 and SIT [42], where UACI values are 14-15% this UACI is 33.637433%.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [42]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given

    The generated custom S-box meets the given criteria.

    “The result is a S-box that meets the given criteria.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Initially, the correlation exceeded 0.9, and after encryption it became approximately zero.

    “Initially, the correlation exceeded 0.9.After encryption, it became approximately zero.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In contrast to [39,41,44], genetic algorithms, including variants and RNA-GA, give NPCR values from 99.4981% to 99.63%, while this method measured NPCR at 99.654677%, the highest among all methods.

    “In contrast to [39,41,44], genetic algorithms, including variants and RNA-GA give NPCR values from 99.4981%, to 99.63%, this method measured NPCR at 99.654677%, the highest among all methods.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [39]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [41]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [44]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Findingown data

    The original images' RGB histograms show peaks and valleys and reflect a non-even spread of pixel intensities.

    “In the images, the RGB histograms show peaks and valleys and the histograms reflect a not even spread of pixel intensities.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The designed S-box affects the histograms and increases the entropy to approximately 8 bits per channel after encryption.

    “It affects the histograms and increases the entropy to approximately 8 bits per channel after encryption.”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data

    The proposed encryption method reduces pixel correlation in horizontal, vertical, and diagonal directions and across all color channels.

    “The method reduces the correlation between pixels.”
    • Also stated in: § 5. 1. Pre/post encryption correlation structure by calculating two-dimensional and three-dimensional correlation coefficients · plain
      “Pixel correlation drops in every direction.”
    • Also stated in: § 5. 1. Pre/post encryption correlation structure by calculating two-dimensional and three-dimensional correlation coefficients · plain
      “Pixel correlation drops across all color channels.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    All fifteen NIST STS categories were passed at α = 0.01, with p-values greater than 0.01.

    “All fifteen NIST STS categories were passed at α = 0.01.The p-values were greater than 0.01.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The proposed method beats RC4, which has NPCR 99.5895% and UACI 14.8831.

    “Beats RC4 has NPCR 99.5895% and UACI 14.8831.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    When compared with ciphers and meta-heuristics, the proposed method gets the NPCR of 99.6547% and a UACI of 33.6374%.

    “When compare the proposed method, with ciphers and meta-heuristics, the proposed method gets the NPCR of 99.6547% and a UACI of 33.6374%.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The proposed method beats DES, which has NPCR 99.5911% and UACI 15.0325.

    “The proposed method beats DES, which has NPCR 99.5911% and UACI 15.0325.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The proposed method slightly beats the heuristics Tabu, which has NPCR 99.6129% and UACI 33.4127, and RNA-GA, which has NPCR 99.6006% and UACI 33.3479.

    “The pro- posed method also slightly beats the heuristics Tabu, which has NPCR 99.6129% and UACI 33.4127 and RNA-GA, which has NPCR 99.6006% and UACI 33.3479.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Operation in CBC mode supplies semantic security.

    “operation in CBC (Cipher Block Chaining) mode supplies semantic security”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: CBC mode gives security when the IVs are pseudorandom and do not repeat.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Performance parity with optimized baseline AES implementations under identical hardware and I/O constraints was not achieved.

    “Second, performance parity with optimized baseline AES implementations under identical hardware and I/O constraints was not achieved.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [5]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Premisereasoning

    FA-based ideas are not presented within a unified framework that integrates formal reversibility, standardized metrics, and benchmarks for comparable systems.

    “However, they are not presented within a unified framework that integrates formal reversibility, standardized metrics, and benchmarks for comparable systems.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Unification of FA-based design principles with a standardized, attack-oriented evaluation beyond aggregate randomness tests was not achieved.

    “First, unification of FA-based design principles with a standardized, attack-oriented evaluation beyond aggregate randomness tests was not achieved.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • The quote only partly states this claim.

    Interpretationreasoning

    Statistical attacks become much harder due to the decorrelation.

    “Statistical attacks become much harder.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Statistically random ciphertext results from replacing ShiftRows and MixColumns in AES with a surjective finite automaton and from using an S-box with high nonlinearity and limited differential homogeneity.

    “Statistically random ciphertext results from two solutions.The first is replacing ShiftRows and MixColumns in AES with a surjective finite automaton.This automaton permutes bits at the state level.The second is using an S-box with high nonlinearity and limited differential homogeneity.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    The NPCR values show that the encryption process is very sensitive and spreads changes well.

    “The NPCR values show that the encryption process is very sensitive and spreads changes well.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    In a spatiotemporal scheme, the diagonal correlation can be around 0.45.

    “For example, the diagonal correlation can be around 0.45 in a spatiotemporal scheme.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationnone given

    The application was demonstrated on RGB images from the USC-SIPI dataset, and generalization to other image types has not yet been performed.

    “One of the limitations of this study is that the application was demonstrated on RGB images from the USC-SIPI dataset.Generalization to other image types has not yet been performed.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    The encryption and decryption scheme is robust against degradation, occlusion, and impulse noise.

    “The results prove that the encryption and decryption scheme is robust, against degradation, occlusion and impulse noise.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The tested occlusion and salt-and-pepper noise levels are representative of real-world data loss and communication channel conditions.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    The custom S-box enhances nonlinearity and reduces correlation.

    “The custom S-box enhances nonlinearity and reduces correlation.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The security gains come from the custom 8-bit S-box that is made to meet limits.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    At present there is no framework that shows advantage over baseline AES in an attacker-focused evaluation.

    “At present there is no framework that shows advantage over baseline AES in an attacker-focused evaluation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The histogram patterns of original images can be used for math attacks or image rebuilding.

    “The histogram patterns match the shapes, surface patterns and color spread in the image and the histogram patterns can be used for math attacks or image rebuilding.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning

    The challenge of combining taxonomy-level ideas with constructive, provably reversible automata capable of scaling without prohibitive growth in the number of states remained open.

    “Furthermore, the challenge of combining taxonomy-level ideas with constructive, provably reversible automata capable of scaling without prohibitive growth in the number of states remained open.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    The high entropy means little information leaks and cipher images resist entropy-based attacks.

    “The high entropy means that little information leaks and that the cipher images resist attacks that look at entropy statistics.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    There is no reversible performance-competitive framework that synthesizes an S-box from finite automata method primitives and guarantees properties such as 8×8 bijectivity, nonlinearity ≥112, differential homogeneity ≤4, algebraic degree ≥7, no linear structures, no fixed points and good boomerang metrics.

    “This study view that there is no reversible performance-competitive framework.The framework must synthesize a S-box from finite automata method (FAM) primitives.The framework must guarantee properties such, as 8 × 8 bijectivity, nonlinearity ≥112, differential homogeneity ≤4 algebraic degree ≥7, no linear structures, no fixed points and good boomerang metrics.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationnone given

    Reproducibility depends on the S-box synthesis, which uses random NP exchanges, and on the construction of the FA random state.

    “Reproducibility depends on the S-box synthesis.The synthesis uses random NP exchanges.It also depends on the construction of the FA random state.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Higher UACI means that even a tiny change in the input image causes a spread of brightness differences, making differential analysis less effective.

    “The higher UACI means that even a tiny change, in the input image causes a spread of brightness differences.Larger spread makes differential analysis less effective.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The surjective finite automaton simulates both row shifts and column mixing.

    “The surjective finite automaton simulates both row shifts and column mixing.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Replacing AES ShiftRows and MixColumns with a finite automaton preserves information-losslessness (decodability) at the machine level and strengthens permutation and diffusion at the bit level.
    • The authors state this as an assumption: The FA is built as a mapping that covers the set of states; each state goes to a state and no information is lost; it works at the bit level; it copies row shifts and column mixing; it is meant to give diffusion then the ShiftRows/MixColumns pair.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    Compressed formats such as JPEG have not been tested.

    “Compressed formats such as JPEG have also not been tested.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    Adding the custom S-box to an AES pipeline makes the encryption scheme more resistant to attacks from methods and new methods.

    “This addition makes the encryption scheme more resistant, to attacks from methods and new methods.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The security gains come from the custom 8-bit S-box that is made to meet limits.

  • May combine two claims.

    Interpretationnone given

    The finite automaton acts as a diffusion layer and breaks the local spatial structure.

    “It acts as a diffusion layer.It breaks the local spatial structure.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The surjective finite automaton maps each state to a state so there is no loss of information.

    “The surjective finite automaton maps each state to a state so there is no loss of information.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Replacing AES ShiftRows and MixColumns with a finite automaton preserves information-losslessness (decodability) at the machine level and strengthens permutation and diffusion at the bit level.
    • The authors state this as an assumption: The FA is built as a mapping that covers the set of states; each state goes to a state and no information is lost; it works at the bit level; it copies row shifts and column mixing; it is meant to give diffusion then the ShiftRows/MixColumns pair.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Study design and conductreasoning

    The USC-SIPI test set used is representative and avoids duplication and overlapping candidates.

    “Its suite is representative.It avoids duplication and excludes overlapping candidates.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The surjective finite automaton works on bit-level data to shuffle and spread the state bits more securely than transformations.

    “The surjective finite automaton works on bit-level data to shuffle and spread the state bits securely than transformations.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Replacing AES ShiftRows and MixColumns with a finite automaton preserves information-losslessness (decodability) at the machine level and strengthens permutation and diffusion at the bit level.
    • The authors state this as an assumption: The FA is built as a mapping that covers the set of states; each state goes to a state and no information is lost; it works at the bit level; it copies row shifts and column mixing; it is meant to give diffusion then the ShiftRows/MixColumns pair.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    The custom S-box provides security and strong linear security.

    “The custom S-box provides security and strong linear security.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The security gains come from the custom 8-bit S-box that is made to meet limits.

  • Classified as probably not a substantial claim.

    Premisecitation

    There is a fundamental gap between statistical tests and proofs targeting specific attack models, such as related-key attacks or quantum algorithms.

    “Furthermore, there is a fundamental gap between statistical tests and proofs targeting specific attack models, such as related-key attacks or quantum algorithms.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    Although the criteria are specified, repeated generation may yield different S-box instances, and the statistics of such instances may differ slightly.

    “Although the criteria are specified, repeated generation may yield different instances.The statistics of such instances may differ slightly.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    At a noise density of 0.2, the PSNR is 29.2 decibels and the image still shows a recognizable structure.

    “At a noise density of 0.2, the PSNR is 29.2 decibels and the image still show a recognizable structure.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Statistical metrics (entropy, chi-square, correlation, NPCR, UACI, PSNR) are sufficient to establish the security and robustness of the proposed image encryption method.

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductnone given

    The security analysis in this paper is based on statistical metrics, robustness is assessed using PSNR, and formal cryptanalysis is not presented.

    “The security analysis in this paper is based on statistical metrics.Robustness is assessed using PSNR.Formal cryptanalysis is not presented.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationnone given

    The use of FA and a custom S-box increases implementation complexity and may create overhead compared to hardware-accelerated AES.

    “The use of FA and a custom S-box increases the implementation complexity.This may create overhead compared to hardware-accelerated AES.”
    Quote in the paperNot a substantial claim
49 of 49

  • The authors state this as an assumption.

    Replacing AES ShiftRows and MixColumns with a finite automaton preserves information-losslessness (decodability) at the machine level and strengthens permutation and diffusion at the bit level.

    “Replace AES ShiftRows and MixColumns with a finite automaton because expect the surjective finite automaton to preserve information-losslessness (decodability) at the machine level and to strengthen permutation and diffusion at the bit level.”

  • The authors state this as an assumption.

    The security gains come from the custom 8-bit S-box that is made to meet limits.

    “The security gains come from the custom 8-bit S-box that is made to meet limits.”

  • The authors state this as an assumption.

    CBC mode gives security when the IVs are pseudorandom and do not repeat.

    “The CBC mode gives security when the IVs are pseudorandom and the IVs do not repeat.”

  • The authors state this as an assumption.

    Statistical metrics and differential metrics such as NIST SP 800-22, NPCR, UACI and robustness tests such as occlusion serve as substitutes for more confusion and a smaller attack surface.

    “The statistical metrics and the differential metrics such, as NIST SP 800-22 NPCR, UACI and the robustness tests such, as occlusion serve as substitutes for more confusion and a smaller attack surface.”

  • The authors state this as an assumption.

    The FA is built as a mapping that covers the set of states; each state goes to a state and no information is lost; it works at the bit level; it copies row shifts and column mixing; it is meant to give diffusion then the ShiftRows/MixColumns pair.

    “It is possible to notice that the FA is built as a mapping that covers the set of states.The FA makes each state go to a state and the FA does not lose any information.The FA works at the bit level.The FA copies the row shifts and the FA copies the column mixing.The FA is meant to give diffusion, then the ShiftRows/MixColumns pair.”

  • The paper relies on this without support.

    The eight standard test images are representative of natural images for evaluating the encryption method.

    “Eight test images were evaluated (”

  • The paper relies on this without support.

    The tested occlusion and salt-and-pepper noise levels are representative of real-world data loss and communication channel conditions.

    “The noisy conditions appear often in communication channels.Examples: networks and damaged data carriers.”

  • The paper relies on this without support.

    Statistical metrics (entropy, chi-square, correlation, NPCR, UACI, PSNR) are sufficient to establish the security and robustness of the proposed image encryption method.

    “The security analysis in this paper is based on statistical metrics.Robustness is assessed using PSNR.Formal cryptanalysis is not presented.”
8 of 8

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Noise placement differs between robustness descriptions

    Major

    One description says salt-and-pepper noise was added to decrypted images; another says it was introduced before decryption. These are different tests, so the text does not consistently identify what the reported robustness results measure.

  • Reported IV depends on ciphertext that requires the IV

    Major

    The stated CBC equation requires the IV to encrypt the first plaintext block. The paper also says that the IV used for the modified-AES tests was computed from the encrypted image, making the described encryption procedure circular without an additional procedure that is not specified.

  • Original-image correlations are not all above 0.9

    Minor

    The narrative says original-image coefficients in every direction and channel exceed 0.9, but the 4.2.06 original row includes several lower values, down to 0.7549.

  • Proposed-method correlation signs disagree with Table 8

    Minor

    The comparison text gives all three proposed-method coefficients as negative, whereas Table 8 prints all three as positive.

  • Figure caption gives a different third crop level

    Minor

    The results describe and tabulate the third crop level as 0.125, while the Fig. 8 caption labels it 0.15.

5 of 5

Cites

Flagged first, then by citations

1 of the 43 references couldn’t be matched to a known paper.

Research works printed but not linked: 1

  1. Salami, Y., Khajevand, V., Zeinali, E. (2023). Cryptographic algorithms: A review of the literature, weaknesses and open challenges. Journal of Computer & Robotics, 2 (16), 63-115. http://dx.doi.org/10.22094/JCR.2023.1983496.1298 · DOI

Authors
Barlybayev, A., Saukhanova, Z., Shakhmetova, G., Sharipbay, A., Raykul, S., & Khassenov, A.
Affiliations
L. N. Gumilyov Eurasian National University (KZ)
ORCID
Venue
Eastern-European Journal of Enterprise Technologies
Publisher
Private Company Technology Center
Published
2025-12-30
Volume
6
Issue
9 (138)
Pages
77-99
ISSN
  • 1729-3774 (ISSN-L)
  • 1729-4061
DOI
10.15587/1729-4061.2025.348368
OpenAlex
W7117781149
Cited by
0
References
43
FWCI
0.00
Type
journal article
Open access
Licence
Locations
Eastern-European Journal of Enterprise Technologies (publishedVersion, cc-by)
Funders
Ministry of Education and Science of the Republic of Kazakhstan
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Chaos-based Image/Signal Encryption
  • Physical Sciences › Computer Science › Artificial Intelligence › Cryptographic Implementations and Security
  • Physical Sciences › Computer Science › Artificial Intelligence › Computer Science and Engineering
Keywords
image encryption, AES-128, CBC mode, diffusion, confusion, NPCR, UACI, statistical security

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-23.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-23.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review