Pith. sign in

REVIEW 4 major objections 6 minor 58 references

Efficient Blockchain-based Steganography via Backcalculating Generative Adversarial Network

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims a reversible GAN can embed 11–24 bits of covert data per Bitcoin transaction amount and up to about 40 bits per numeric field under IEEE 754 precision, with the receiver recovering the data by inverting the generator.

desk verdict Novel reversible-GAN steganography with plausible capacity numbers, but the printed decoding equations don't invert the generator as written. read the letter →

arxiv 2506.16023 v1 pith:ILDEVTNX submitted 2025-06-19 cs.CR

Zhuo Chen , Jialing He , Jiacheng Wang , Zehui Xiong , Tao Xiang , Liehuang Zhu , Dusit Niyato This is my paper · ORCID
classification cs.CR
keywords blockchainsteganographycovertchannelreversiblegenerativeadversarialnetworktransactionamountembeddingcapacityexpansionIEEE754precisionsteganalysis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to turn the required fields of a blockchain transaction—fields the sender must fill anyway, like the output amount and the fee—into a covert channel. The sender trains a generative adversarial network whose generator is deliberately invertible, encodes extra covert data into the generator's input noise, and publishes the generated field; the receiver recovers the noise by backcalculating the generator. The central experimental claim is that this works on real Bitcoin data: 11 bits per transaction amount with the basic design, 24 bits with an improved design that reduces rounding error, both with generation times under three seconds. The paper also claims a hard ceiling: under IEEE 754 double precision, about 40 bits per numeric field, because floating-point error dominates once dataset magnitude reaches $10^{17}$. A reader should care because this raises channel capacity without touching the fields where existing steganography already hides data.

What carries the argument

The load-bearing object is the reversible generator: two fully connected layers, equal width M=H=N, with LeakyReLU followed by Sigmoid (or ClipSigmoid). Decoding is the exact inversion chain X̂ = $W1^{{-1}}$($LeakyReLU^{{-1}}$($W2^{{-1}}$(Logistic(Ŷ)))), so invertibility of W1 and W2 is assumed. The argument that capacity can be tuned is carried by the NPID measure—the count of consecutive identical leading digits between the sender's normalized value and the receiver's rounded value—and by the theorem that raising the dataset's maximum value by a factor of ten increases NPID by one.

What would settle it

Train the R-GAN exactly as described on Bitcoin output amounts, then compute the condition number of W1 and W2 from the trained generator and run the encoding/decoding loop on 10,000 random noise vectors. If any trained run yields a singular or severely ill-conditioned matrix, or if the recovered covert-data bits disagree with the sent bits at a rate higher than the reported 11-bit (R-GAN) or 24-bit (CCR-GAN) error rate, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a GAN generator made of two fully connected layers with piecewise-linear and monotone activations is a reversible function, so the sender can put covert data into the generator's noise and the receiver can recover it by computing $W1^{{-1}}$, the inverse LeakyReLU, $W2^{{-1}}$, and the inverse logistic on the observed rounded field. Rounding to integer amounts creates the limiting error; the paper's Theorem 1 says each tenfold increase in the difference between the dataset's maximum and minimum adds one recoverable digit, which motivates keeping extreme amounts in the training set. To make that work, ClipSigmoid clamps tiny Sigmoid outputs to $10^{-20}$ so gradients vanish instead of exploding. The net experimental result is that R-GAN reaches 11 bits and CCR-GAN 24 bits per Bitcoin amount, with the Bitcoin fee, Ethereum amount, and Ethereum fee also usable, and with capacity saturating near 40 bits when the rounding error no longer dominates.

Load-bearing premise

The decoding scheme assumes the generator's two square weight matrices remain full-rank and numerically stable after training; the paper sets their dimensions equal but never checks that the learned matrices are actually invertible.

Editorial extensions

If this is right

  • Existing blockchain covert channels that use the address or signature as the embedding field can gain capacity by also using the amount or fee field, since the sender controls these fields anyway.
  • The capacity ceiling is architectural: with float64 and IEEE 754, no amount of dataset preprocessing pushes a single numeric field beyond about 40 bits.
  • The scheme transfers to other numeric fields and blockchains, with measured capacities of 2–4 bits for Bitcoin fee, roughly 40 bits for Ethereum amount, and 21–25 bits for Ethereum fee.
  • T2C gives the communicating parties a tunable knob: shrinking the dataset magnitude by a factor of ten trades 2–3 bits of capacity for 3–4% better concealment.
  • Concealment is probabilistic, not perfect: a steganalysis classifier trained to recognize generated fields detects them with accuracy between 0.577 and 0.968 depending on field and dataset.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because decodability rests on inverting W1 and W2, a natural testable extension is to regularize the generator to keep its weight matrices well-conditioned; the paper does not verify conditioning after training.
  • The verification loop in the encoder is a rejection sampler whose expected runtime grows roughly exponentially with m; the reported 'under 3 seconds' figure is tied to the tested bit counts and would grow sharply if m were raised, unless error-correcting codes replace bit-by-bit verification.
  • The 40-bit ceiling suggests the true carrier is the field's numeric range, not the neural network; any generator that maps noise to the same distribution would hit the same rounding and precision limits.
  • The paper assumes both parties share the same trained model; a deployment would need a retraining and sync schedule, and that schedule itself could leak side information to an observer.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes a generic blockchain-based steganography framework (GBSF) in which the sender generates required transaction fields (e.g., Bitcoin amount, fees) using a GAN whose generator is made reversible, and embeds additional covert data into the generator's input noise. The receiver decodes the data by inverting the generator. The authors present two schemes, R-GAN and CCR-GAN, the latter adding a counter-intuitive data preprocessing (CIDP) step and a custom ClipSigmoid activation to reduce rounding error and stabilize training. They also propose T2C, a magnitude-reduction mechanism to trade capacity for concealment. Experiments on real Bitcoin and Ethereum amount/fee datasets report per-field capacities of 11 and 24 bits for R-GAN and CCR-GAN on Bitcoin amounts, up to 40-41 bits on Ethereum amounts, and concealment metrics based on the CTR steganalysis model. The central claims are that the proposed schemes improve channel capacity over existing blockchain steganography baselines while maintaining acceptable concealment, and that capacity is ultimately limited by IEEE 754 floating-point precision to about 40 bits per field.

Significance. If the reversibility mechanism is correct, the paper addresses a genuinely under-explored problem: generating required transaction fields while simultaneously embedding covert data, rather than only encoding into existing fields. The empirical evaluation is direct and useful: capacities are measured by round-trip decoding through the actual trained models, and the scaling experiments across Bitcoin and Ethereum amount/fee fields provide a concrete, falsifiable statement about when precision limits capacity. The T2C trade-off is a practical contribution, and the comparison with prior amount-embedding baselines is informative. However, the central decoding procedure as printed is not the mathematical inverse of the described generator, and the proof of the main capacity-scaling theorem is incomplete. These are load-bearing issues because the claimed 11/24/40-bit capacities depend on exact inversion and on Theorem 1's predictions. With a corrected and reproducible decoding specification and a more rigorous theorem, the work would be a solid contribution to blockchain steganography.

major comments (4)
  1. [II.E, Eqs. (6)-(9)] The decoding procedure as printed does not invert the generator. The fully connected layers in PyTorch compute z = W x + b, but steps (3) and (5) use only W2^{-1} and W1^{-1} without subtracting the bias vectors b2 and b1. Moreover, step (2) applies the logistic (sigmoid) function to the normalized amount, whereas inverting the generator's final sigmoid requires the logit function. Unless the Linear layers are explicitly bias-free (the paper never states this, and Fig. 4 labels them only as 'Linear') and Eq. (6) is intended as the inverse sigmoid, exact arithmetic decoding fails regardless of rounding or precision error. Since the receiver's recovered bits are the central measured quantity, this is load-bearing. The paper should state explicitly that all linear layers are bias-free, or add the bias-subtraction steps, and replace Eq. (6) with the logit transform; alternatively, the authors should provide code confirming the exact implemented inverse.
  2. [II.C and II.E] The scheme requires the square weight matrices W1 and W2 to be invertible, but the paper only sets M=H=N and notes that W^{-1} exists if the matrix is full-rank. Standard GAN training with random initialization and gradient descent does not guarantee full rank, nor does it control the condition number. If either matrix is singular or ill-conditioned, the decoded noise is garbage or the error is amplified, which would invalidate the measured m values for the receiver. Please report condition numbers or singular value distributions for the trained weight matrices, or add an explicit invertibility/orthogonality regularization to the generator, and verify that the inverse used in decoding is the true inverse of the trained forward layer.
  3. [III.A, Theorem 1 proof] The proof of Theorem 1 covers only the special case where Q = max(X) is an exact power of ten (Q = 10^m), y has a finite decimal expansion, and the rounding operator [.] behaves as rounding down (y_{m+1} <= 4). In the actual datasets Q is an arbitrary integer (e.g., 2,874,993,345,277 for the Bitcoin amount dataset), rounding is to the nearest integer, and the conclusion NPID(y, \hat{y}) = m+1 does not follow from the stated assumptions for general Q. Since Theorem 1 is used to justify CIDP and to predict the 10^17 saturation observed in Section V.C, this is not a purely cosmetic gap. The theorem should be restated and proved for arbitrary integer Q and the actual rounding mode, or its predictions should be supported by direct empirical validation across the fields used in the paper.
  4. [III.B and II.E] ClipSigmoid as defined in Eq. (19) has a flat segment where it is not invertible. The paper asserts that this does not affect reversibility because the sender can intentionally train the generator to avoid the zero-segment, but no mechanism or experiment demonstrates that the flat segment is never entered for the adopted values of m and the tested datasets. The receiver's decoding equations contain no rule for handling outputs at the ClipSigmoid threshold. If a generated amount corresponds to the flat segment, the receiver cannot recover the sender's pre-activation, and the verification loop in Algorithm 1 would be validating against an inverse that is not the true inverse of the generator. Please provide a more precise argument or experiments showing that the flat segment is avoided for the reported configurations, or modify ClipSigmoid to be bijective.
minor comments (6)
  1. [Figs. 3(a) and 9] The text of the figures contains garbled glyph sequences such as '/uni00000016/uni00000017/...', making the figures illegible in the submitted version; they should be regenerated with proper fonts.
  2. [II.E, Eq. (6)] The variable name 'Logistic' is used for the function that should invert the sigmoid; please clarify that this is the logit (inverse sigmoid) transformation, or use a different symbol to avoid ambiguity.
  3. [IV, Algorithm 3] The comment above Algorithm 3 states 'Algorithm 2 describes the process of revovering magnitude', but the algorithm being described is Algorithm 3; there is also a typo 'revovering'.
  4. [V.C] There are several typos: 'The Ethernet fee' should be 'Ethereum fee', 'Futhermore' should be 'Furthermore', and 'revovering' appears again.
  5. [V.F, Table IV] Table IV does not state which m values are used for the Ethereum amount row after the paper discards CCR-GAN and selects a reduced magnitude of 10^3; please make the mapping to Table III explicit.
  6. [V.D, Answer to RQ5] The quantitative trade-off 'for every 10-fold reduction in dataset magnitude, the capacity is reduced by 2-3 bits and concealment is enhanced by 3%-4%' is stated without a dedicated experiment or a direct pointer to the supporting data in Table III; please provide the evidence or clarify how the numbers were obtained.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; capacity claims are measured and the scaling law is arithmetic, with only minor non-load-bearing self-citation in the CTR benchmark.

full rationale

The derivation chain is largely self-contained. The headline capacities (11/24 bits for Bitcoin amount, 40-bit IEEE-754 ceiling) are not obtained by fitting a model and then predicting the same quantity; they are measured from the round-trip verification procedure in Algorithm 1 (m = min NPID2(x̂_i, x_i) - 1) with an explicit time budget in Section V-B and Table I. Theorem 1 is an arithmetic statement about decimal rounding after min-max normalization: combining equations (11)-(13) gives ŷ = floor(yQ)/Q, so a ten-fold increase in Q preserves one more digit directly from the rounding arithmetic; it does not assume the experimental m values. The 40-bit ceiling is validated by controlled magnitude-reduction experiments in Table II rather than being read off the target. The concealment metric uses the CTR model of [15]; since J. He and L. Zhu are co-authors of both works this is a self-citation, but the CTR detector is a fixed published steganalysis benchmark external to the fitted values, so it does not make the central claim circular. An in-scope non-circular weakness: Section II.E Eqs. (7) and (9) invert the fully connected layers without subtracting bias terms, so the printed inverse is not the inverse of the generator as written unless the layers are bias-free; this is a correctness and reproducibility caveat, not a circularity.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central scheme rests on several assumptions that are plausible but not established by the paper: trained square weight matrices stay invertible and stable, sender and receiver share the exact model and normalization constants, blockchain fields can be set to arbitrary generated integers, and the rounding-error scaling in Theorem 1 holds for the relevant cases. The payload size m is measured empirically per dataset rather than predicted, and the ClipSigmoid threshold and T2C exponent are hand-set hyperparameters.

free parameters (5)
  • payload size m = 11 (R-GAN BTC amount), 24 (CCR-GAN BTC amount), 41 (R-GAN ETH amount), etc.
    Chosen per dataset by measuring NPID of round-tripped noise and a time budget; not derived from first principles.
  • ClipSigmoid threshold = 1e-20
    Set by experience as the square of the quotient of min and max dataset values; controls gradient suppression and convergence.
  • T2C reduced magnitude exponent lambda = varies (e.g., 10^3 reduction for ETH amount comparison)
    User-tunable knob trading capacity for concealment; the distribution of trailing lambda digits is estimated from data.
  • R-GAN training data filter = amounts with 5-7 digits (about 10^4 to 10^7)
    Only the most frequent length range is kept for R-GAN, affecting the normalized distribution and the reported capacity.
  • LeakyReLU slope alpha = 0.3
    Hyperparameter affecting invertibility and decoding error.
assumptions (6)
  • domain assumption Square weight matrices W1 and W2 are full-rank after training, so their inverses exist.
    Section II.E requires W2^{-1} and W1^{-1}; training does not enforce full rank.
  • domain assumption Sigmoid, LeakyReLU and ClipSigmoid are invertible on the relevant output range.
    Decoding applies logit and inverse LeakyReLU; the zero-gradient segment of ClipSigmoid is avoided by assuming trained outputs stay outside it.
  • domain assumption Blockchain amount and fee fields are integers fully controlled by the sender and can be set to generated values.
    Section II.C chooses the output amount as the expansion field; protocol constraints such as dust, input balance, and fee consistency are not modeled.
  • domain assumption Sender and receiver can obtain identical trained models and normalization constants.
    Section II.B defers model synchronization to prior work [29,30] with shared seeds every 3 days; exact synchronization is not implemented.
  • ad hoc to paper Theorem 1's rounding behavior assumes a finite decimal expansion and only the rounding-down case (y_{m+1} <= 4).
    Equations (16)-(17) assume the truncated digit is at most 4 and a finite expansion; round-up cases and infinite expansions are ignored.
  • domain assumption IEEE 754 float64 provides about 17 decimal digits, bounding recoverable bits at roughly 40.
    Used to explain the capacity ceiling in Table II; the exact 40-bit bound is empirical, not derived from the standard.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Blockchain-based Steganography via Backcalculating Generative Adversarial Network." pith.science (2026). https://pith.science/paper/ILDEVTNX

@misc{pith2026250616023,
  author       = {Pith},
  title        = {Pith review of: Efficient Blockchain-based Steganography via Backcalculating Generative Adversarial Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ILDEVTNX}},
  note         = {Machine review of arXiv:2506.16023}
}
read the original abstract

Blockchain-based steganography enables data hiding via encoding the covert data into a specific blockchain transaction field. However, previous works focus on the specific field-embedding methods while lacking a consideration on required field-generation embedding. In this paper, we propose a generic blockchain-based steganography framework (GBSF). The sender generates the required fields such as amount and fees, where the additional covert data is embedded to enhance the channel capacity. Based on GBSF, we design a reversible generative adversarial network (R-GAN) that utilizes the generative adversarial network with a reversible generator to generate the required fields and encode additional covert data into the input noise of the reversible generator. We then explore the performance flaw of R-GAN. To further improve the performance, we propose R-GAN with Counter-intuitive data preprocessing and Custom activation functions, namely CCR-GAN. The counter-intuitive data preprocessing (CIDP) mechanism is used to reduce decoding errors in covert data, while it incurs gradient explosion for model convergence. The custom activation function named ClipSigmoid is devised to overcome the problem. Theoretical justification for CIDP and ClipSigmoid is also provided. We also develop a mechanism named T2C, which balances capacity and concealment. We conduct experiments using the transaction amount of the Bitcoin mainnet as the required field to verify the feasibility. We then apply the proposed schemes to other transaction fields and blockchains to demonstrate the scalability. Finally, we evaluate capacity and concealment for various blockchains and transaction fields and explore the trade-off between capacity and concealment. The results demonstrate that R-GAN and CCR-GAN are able to enhance the channel capacity effectively and outperform state-of-the-art works.

Figures

Figures reproduced from arXiv: 2506.16023 by the authors.

Figure 1
Figure 1. Overview of GBSF framework. In GBSF, the sender uses the R-GAN to generate the required fields and complete a [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. General workflow of the expanding covert channel. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Example of Bitcoin transaction amount. The [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Overview of R-GAN. U refers to a uniform distribution; B is the batchsize; H represents the hidden dimension; M denotes the input dimension; N is the output dimension. BCELoss refers to the binary cross entropy loss [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 7
Figure 7. Figure 7: Sigmoid-related weight update process. W represents any of the weight parameters in the model before the second fully connected layer, x denotes the output of the second fully connected layer, y is the output of Sigmoid, and J refers to the loss function. B. ClipSigmoi…
Figure 9
Figure 9. Figure 9: Experiments on recovery bits (including 1-bit MSB). [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Time overhead to find a satisfying noise vector. [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: The effect of ClipSigmoid. Sigmoid and ClipSigmoid as the activation function during the training [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 57 canonical work pages

  1. [1]

    Digital image steganography: A literature survey,

    P. C. Mandal, I. Mukherjee, G. Paul, and B. Chatterji, “Digital image steganography: A literature survey,”In- formation Sciences, 2022

    work page 2022

  2. [2]

    When deep learning meets steganogra- phy: Protecting inference privacy in the dark,

    Q. Liu, J. Yang, H. Jiang, J. Wu, T. Peng, T. Wang, and G. Wang, “When deep learning meets steganogra- phy: Protecting inference privacy in the dark,” inIEEE INFOCOM 2022-IEEE Conference on Computer Com- munications. IEEE, 2022, pp. 590–599

    work page 2022

  3. [3]

    Wmdefense: Using watermark to defense byzantine attacks in federated learning,

    X. Zheng, Q. Dong, and A. Fu, “Wmdefense: Using watermark to defense byzantine attacks in federated learning,” inIEEE INFOCOM 2022-IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS). IEEE, 2022, pp. 1–6

    work page 2022

  4. [4]

    Natural language watermarking via paraphraser-based lexical substitution,

    J. Qiang, S. Zhu, Y . Li, Y . Zhu, Y . Yuan, and X. Wu, “Natural language watermarking via paraphraser-based lexical substitution,”Artificial Intelligence, vol. 317, p. 103859, 2023

    work page 2023

  5. [5]

    Balboa: Bobbing and weaving around network censorship,

    M. B. Rosen, J. Parker, and A. J. Malozemoff, “Balboa: Bobbing and weaving around network censorship,” in 30th USENIX Security Symposium (USENIX Security 21), 2021, pp. 3399–3413

    work page 2021

  6. [6]

    Generative steganography via auto- generation of semantic object contours,

    Z. Zhou, X. Dong, R. Meng, M. Wang, H. Yan, K. Yu, and K.-K. R. Choo, “Generative steganography via auto- generation of semantic object contours,”IEEE Transac- tions on Information Forensics and Security, 2023

    work page 2023

  7. [7]

    Robust invertible image steganography,

    Y . Xu, C. Mou, Y . Hu, J. Xie, and J. Zhang, “Robust invertible image steganography,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 7875–7884

    work page 2022

  8. [8]

    Generative ai for integrated sensing and communication: Insights from the physical layer perspective,

    J. Wang, H. Du, D. Niyato, J. Kang, S. Cui, X. S. Shen, and P. Zhang, “Generative ai for integrated sensing and communication: Insights from the physical layer perspective,”IEEE Wireless Communications, 2024

    work page 2024

Show all 58 references
  1. [9]

    Covert channels in blockchain and blockchain based covert communication: Overview, state-of-the-art, and future directions,

    T. Zhang, B. Li, Y . Zhu, T. Han, and Q. Wu, “Covert channels in blockchain and blockchain based covert communication: Overview, state-of-the-art, and future directions,”Computer Communications, 2023

    2023

  2. [10]

    Privacy-preserving byzantine-robust federated learning via blockchain systems,

    Y . Miao, Z. Liu, H. Li, K.-K. R. Choo, and R. H. Deng, “Privacy-preserving byzantine-robust federated learning via blockchain systems,”IEEE Transactions on Infor- mation Forensics and Security, vol. 17, pp. 2848–2861, 2022. 16

    2022

  3. [11]

    Blockchain meets covert communication: A survey,

    Z. Chen, L. Zhu, P. Jiang, C. Zhang, F. Gao, J. He, D. Xu, and Y . Zhang, “Blockchain meets covert communication: A survey,”IEEE Communications Surveys & Tutorials, 2022

    2022

  4. [12]

    The applications of blockchain in the covert communication,

    B. Du, D. He, M. Luo, C. Peng, and Q. Feng, “The applications of blockchain in the covert communication,” Wireless Communications and Mobile Computing, vol. 2022, 2022

    2022

  5. [13]

    A covert channel over blockchain based on label tree without long waiting times,

    Z. Wang, L. Zhang, R. Guo, G. Wang, J. Qiu, S. Su, Y . Liu, G. Xu, and Z. Tian, “A covert channel over blockchain based on label tree without long waiting times,”Computer Networks, p. 109843, 2023

    2023

  6. [14]

    Ebdl: Effective blockchain-based covert storage channel with dynamic labels,

    C. Zhang, L. Zhu, C. Xu, Z. Zhang, and R. Lu, “Ebdl: Effective blockchain-based covert storage channel with dynamic labels,”Journal of Network and Computer Ap- plications, vol. 210, p. 103541, 2023

    2023

  7. [15]

    Practical blockchain-based steganographic com- munication via adversarial ai: A case study in bitcoin,

    M. Wang, Z. Zhang, J. He, F. Gao, M. Li, S. Xu, and L. Zhu, “Practical blockchain-based steganographic com- munication via adversarial ai: A case study in bitcoin,” The Computer Journal, vol. 65, no. 11, pp. 2926–2938, 2022

    2022

  8. [16]

    Generative adversarial networks: An overview,

    A. Creswell, T. White, V . Dumoulin, K. Arulkumaran, B. Sengupta, and A. A. Bharath, “Generative adversarial networks: An overview,”IEEE signal processing maga- zine, vol. 35, no. 1, pp. 53–65, 2018

    2018

  9. [17]

    Generative ai based se- cure wireless sensing for isac networks,

    J. Wang, H. Du, Y . Liu, G. Sun, D. Niyato, S. Mao, D. I. Kim, and X. Shen, “Generative ai based se- cure wireless sensing for isac networks,”arXiv preprint arXiv:2408.11398, 2024

    2024 arXiv

  10. [18]

    An extension of imagechain concept that allows multiple images per block,

    K. Koptyra and M. R. Ogiela, “An extension of imagechain concept that allows multiple images per block,” inIEEE INFOCOM 2022-IEEE Conference on Computer Communications Workshops (INFOCOM WK- SHPS). IEEE, 2022, pp. 1–2

    2022

  11. [19]

    Magview: A distributed magnetic covert channel via video encoding and decoding,

    J. Zhang, X. Ji, W. Xu, Y .-C. Chen, Y . Tang, and G. Qu, “Magview: A distributed magnetic covert channel via video encoding and decoding,” inIEEE INFOCOM 2020-IEEE Conference on Computer Communications. IEEE, 2020, pp. 357–366

    2020

  12. [20]

    Semantic-preserving linguistic steganography by pivot translation and semantic-aware bins coding,

    T. Yang, H. Wu, B. Yi, G. Feng, and X. Zhang, “Semantic-preserving linguistic steganography by pivot translation and semantic-aware bins coding,”IEEE Transactions on Dependable and Secure Computing, 2023

    2023

  13. [21]

    Lin- guistic steganalysis toward social network,

    J. Yang, Z. Yang, J. Zou, H. Tu, and Y . Huang, “Lin- guistic steganalysis toward social network,”IEEE Trans- actions on Information Forensics and Security, vol. 18, pp. 859–871, 2022

    2022

  14. [22]

    Gasla: Enhancing the applicability of sign language translation,

    J. Li, Y . Liu, W. Xu, and Z. Li, “Gasla: Enhancing the applicability of sign language translation,” inIEEE INFOCOM 2022-IEEE Conference on Computer Com- munications. IEEE, 2022, pp. 1249–1258

    2022

  15. [23]

    Unseencode: In- visible on-screen barcode with image-based extraction,

    H. Cui, H. Bian, W. Zhang, and N. Yu, “Unseencode: In- visible on-screen barcode with image-based extraction,” inIEEE INFOCOM 2019-IEEE Conference on Computer Communications. IEEE, 2019, pp. 1315–1323

    2019

  16. [24]

    Preventing non-intrusive load mon- itoring privacy invasion: A precise adversarial attack scheme for networked smart meters,

    J. He, J. Wang, N. Wang, S. Guo, L. Zhu, D. Niy- ato, and T. Xiang, “Preventing non-intrusive load mon- itoring privacy invasion: A precise adversarial attack scheme for networked smart meters,”arXiv preprint arXiv:2412.16893, 2024

    2024 arXiv

  17. [25]

    Advancing non-intrusive load monitoring: Pre- dicting appliance-level power consumption with indirect supervision,

    J. He, J. Feng, S. Guo, Z. Chen, Y . Liu, T. Xiang, and L. Zhu, “Advancing non-intrusive load monitoring: Pre- dicting appliance-level power consumption with indirect supervision,”IEEE Transactions on Network Science and Engineering, 2025

    2025

  18. [26]

    Single-precision floating-point adder based on ieee 754 standard through verilog,

    C. Guo, Z. Jia, Y . Lin, and H. Yang, “Single-precision floating-point adder based on ieee 754 standard through verilog,” inInternational Conference on Artificial Intelli- gence and Communication Technology. Springer, 2024, pp. 141–150

    2024

  19. [27]

    Blockchain-based group covert communication for iot network,

    X. Yuan, P. Jiang, Z. Chen, C. Zhang, F. Gao, and L. Zhu, “Blockchain-based group covert communication for iot network,”IEEE Internet of Things Journal, 2025

    2025

  20. [28]

    Tackling data mining risks: A tripartite covert channel merging blockchain and ipfs,

    Z. Chen, L. Zhu, P. Jiang, J. He, and Z. Zhang, “Tackling data mining risks: A tripartite covert channel merging blockchain and ipfs,”IEEE Transactions on Network Science and Engineering, 2025

    2025

  21. [29]

    Secret-to-image reversible transformation for generative steganography,

    Z. Zhou, Y . Su, J. Li, K. Yu, Q. J. Wu, Z. Fu, and Y . Shi, “Secret-to-image reversible transformation for generative steganography,”IEEE Transactions on Dependable and Secure Computing, vol. 20, no. 5, pp. 4118–4134, 2022

    2022

  22. [30]

    Provably secure disambiguating neural linguistic steganography,

    Y . Qi, K. Chen, K. Zeng, W. Zhang, and N. Yu, “Provably secure disambiguating neural linguistic steganography,” IEEE Transactions on Dependable and Secure Comput- ing, 2024

    2024

  23. [31]

    Katz and Y

    J. Katz and Y . Lindell,Introduction to modern cryptogra- phy: principles and protocols. Chapman and hall/CRC, 2007

    2007

  24. [32]

    Errorless robust jpeg steganography using outputs of jpeg coders,

    J. Butora, P. Puteaux, and P. Bas, “Errorless robust jpeg steganography using outputs of jpeg coders,”IEEE Transactions on Dependable and Secure Computing, vol. 21, no. 4, pp. 2394–2406, 2023

    2023

  25. [33]

    Accessed: June 8,

    Pytorch versions installation. Accessed: June 8,

  26. [34]

    Chain-based covert data embedding schemes in blockchain,

    H. Cao, H. Yin, F. Gao, Z. Zhang, B. Khoussainov, S. Xu, and L. Zhu, “Chain-based covert data embedding schemes in blockchain,”IEEE Internet of Things Journal, vol. 9, no. 16, pp. 14 699–14 707, 2020

    2020

  27. [35]

    Exploring covert channels in bitcoin trans- actions,

    A. Fionov, “Exploring covert channels in bitcoin trans- actions,” in2019 International Multi-Conference on En- gineering, Computer and Information Sciences (SIBIR- CON). IEEE, 2019, pp. 0059–0064

    2019

  28. [36]

    Analysis of arbitrary content on blockchain-based systems using bigquery,

    M. Gregoriadis, R. Muth, and M. Florian, “Analysis of arbitrary content on blockchain-based systems using bigquery,” inCompanion Proceedings of the Web Con- ference 2022, 2022, pp. 478–487

    2022

  29. [37]

    Dlchain: A covert channel over blockchain based on dynamic labels,

    J. Tian, G. Gou, C. Liu, Y . Chen, G. Xiong, and Z. Li, “Dlchain: A covert channel over blockchain based on dynamic labels,” inInformation and Communications Security: 21st International Conference, ICICS 2019, Beijing, China, December 15–17, 2019, Revised Selected Papers 21. ...

    2019

  30. [38]

    A novel covert communication method based on bitcoin 17 transaction,

    X. Luo, P. Zhang, M. Zhang, H. Li, and Q. Cheng, “A novel covert communication method based on bitcoin 17 transaction,”IEEE Transactions on Industrial Informat- ics, vol. 18, no. 4, pp. 2830–2839, 2021

    2021

  31. [39]

    A blockchain-based secure covert communication method via shamir threshold and stc mapping,

    P. Zhang, Q. Cheng, M. Zhang, and X. Luo, “A blockchain-based secure covert communication method via shamir threshold and stc mapping,”IEEE Transac- tions on Dependable and Secure Computing, 2024

    2024

  32. [40]

    Amount-based covert communication over blockchain,

    Y . Tian, X. Liao, L. Dong, Y . Xu, and H. Jiang, “Amount-based covert communication over blockchain,” IEEE Transactions on Network and Service Management, 2024

    2024

  33. [41]

    Accessed: June 8, 2025

    Pay-to-pubkey hash. Accessed: June 8, 2025. [Online]. Available: http://bitcoinwiki.org/wiki/pay-to- pubkey-hash/

    2025

  34. [42]

    Accessed: June 1, 2023

    Bitcoin protocol layer statistics. Accessed: June 1, 2023. [Online]. Available: https://transactionfee.info/

    2023

  35. [43]

    Provably secure covert communication on blockchain,

    J. Partala, “Provably secure covert communication on blockchain,”Cryptography, vol. 2, no. 3, p. 18, 2018

    2018

  36. [44]

    Achieving a covert channel over an open blockchain network,

    F. Gao, L. Zhu, K. Gai, C. Zhang, and S. Liu, “Achieving a covert channel over an open blockchain network,”IEEE Network, vol. 34, no. 2, pp. 6–13, 2020

    2020

  37. [45]

    Research on a covert communication model realized by using smart contracts in blockchain environment,

    L. Zhang, Z. Zhang, W. Wang, Z. Jin, Y . Su, and H. Chen, “Research on a covert communication model realized by using smart contracts in blockchain environment,”IEEE Systems Journal, vol. 16, no. 2, pp. 2822–2833, 2021

    2021

  38. [46]

    An approach of covert communication based on the ethereum whisper protocol in blockchain,

    L. Zhang, Z. Zhang, Z. Jin, Y . Su, and Z. Wang, “An approach of covert communication based on the ethereum whisper protocol in blockchain,”International Journal of Intelligent Systems, vol. 36, no. 2, pp. 962–996, 2021

    2021

  39. [47]

    Uncontrolled randomness in blockchains: Covert bulletin board for illicit activity,

    N. Alsalami and B. Zhang, “Uncontrolled randomness in blockchains: Covert bulletin board for illicit activity,” in2020 IEEE/ACM 28th International Symposium on Quality of Service (IWQoS). IEEE, 2020, pp. 1–10

    2020

  40. [48]

    Research on information hiding methods based on blockchain technology,

    J. Liu, “Research on information hiding methods based on blockchain technology,” Master’s thesis, Nanjing Uni- versity of Information Science and Technology, 2020

    2020

  41. [49]

    Lsc: Online auto-update smart contracts for fortifying blockchain-based log systems,

    W. Shao, Z. Wang, X. Wang, K. Qiu, C. Jia, and C. Jiang, “Lsc: Online auto-update smart contracts for fortifying blockchain-based log systems,”Information Sciences, vol. 512, pp. 506–517, 2020

    2020

  42. [50]

    Auto-generation of smart contracts from a domain-specific xml-based language,

    V . Dwivedi and A. Norta, “Auto-generation of smart contracts from a domain-specific xml-based language,” in Intelligent Data Engineering and Analytics: Proceedings of the 9th International Conference on Frontiers in Intelligent Computing: Theory and Applications (FICTA 2021)....

    2021

  43. [51]

    Designing secure ethereum smart contracts: A finite state machine based approach,

    A. Mavridou and A. Laszka, “Designing secure ethereum smart contracts: A finite state machine based approach,” inFinancial Cryptography and Data Security: 22nd In- ternational Conference, FC 2018, Nieuwpoort, Curac ¸ao, February 26–March 2, 2018, Revised Selected Papers 22. Sp...

    2018

  44. [52]

    Scsguard: Deep scam detec- tion for ethereum smart contracts,

    H. Hu, Q. Bai, and Y . Xu, “Scsguard: Deep scam detec- tion for ethereum smart contracts,” inIEEE INFOCOM 2022-IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS). IEEE, 2022, pp. 1–6

    2022

  45. [53]

    Msdc: Exploiting multi-state power consumption in non-intrusive load monitoring based on a dual-cnn model,

    J. He, J. Liu, Z. Zhang, Y . Chen, Y . Liu, B. Khous- sainov, and L. Zhu, “Msdc: Exploiting multi-state power consumption in non-intrusive load monitoring based on a dual-cnn model,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 4, pp. 5078–5086, 2023

    2023

  46. [54]

    Vae-stega: Linguistic steganography based on varia- tional auto-encoder,

    Z. Yang, S. Zhang, Y . Hu, Z. Hu, and Y . Huang, “Vae-stega: Linguistic steganography based on varia- tional auto-encoder,”IEEE Transactions on Information Forensics and Security, vol. 16, 2021

    2021

  47. [55]

    Linguistic generative steganography with enhanced cognitive-imperceptibility,

    Z. Yang, L. Xiang, S. Zhang, X. Sun, and Y . Huang, “Linguistic generative steganography with enhanced cognitive-imperceptibility,”IEEE Signal Processing Let- ters, vol. 28, pp. 409–413, 2021

    2021

  48. [56]

    Topic-aware neural linguistic steganography based on knowledge graphs,

    Y . Li, J. Zhang, Z. Yang, and R. Zhang, “Topic-aware neural linguistic steganography based on knowledge graphs,”ACM/IMS Transactions on Data Science, vol. 2, no. 2, pp. 1–13, 2021

    2021

  49. [57]

    Linguistic steganography based on adaptive probability distribution,

    X. Zhou, W. Peng, B. Yang, J. Wen, Y . Xue, and P. Zhong, “Linguistic steganography based on adaptive probability distribution,”IEEE Transactions on Depend- able and Secure Computing, vol. 19, no. 5, pp. 2982– 2997, 2021. Zhuo Chenreceived the B.E. degree in informa- tion secu...

    2021

  50. [2025]

    Available: https://pytorch.org/get- started/previous-versions/

    [Online]. Available: https://pytorch.org/get- started/previous-versions/

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.