Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography
Open accessThe DOI matches its Crossref record; 20 of the 20 references it cites were found in a registry or an index; 3 of the 6 sources its claims cite were checked for support.
Abstract
Researches in Artificial Intelligence (AI) have achieved many important breakthroughs, especially in recent years. In some cases, AI learns alone from scratch and performs human tasks faster and better than humans. With the recent advances in AI, it is natural to wonder whether Artificial Neural Networks will be used to successfully create or break cryptographic algorithms. Bibliographic review shows the main approach to this problem have been addressed throughout complex Neural Networks, but without understanding or proving the security of the generated model. This paper presents an analysis of the security of cryptographic algorithms generated by a new technique called Adversarial Neural Cryptography (ANC). Using the proposed network, we show limitations and directions to improve the current approach of ANC. Training the proposed Artificial Neural Network with the improved model of ANC, we show that artificially intelligent agents can learn the unbreakable One-Time Pad (OTP) algorithm, without human knowledge, to communicate securely through an insecure communication channel. This paper shows in which conditions an AI agent can learn a secure encryption scheme. However, it also shows that, without a stronger adversary, it is more likely to obtain an insecure one.
Overview
CoChat Summary
TL;DR
The paper analyzes Adversarial Neural Cryptography (ANC) and shows that the original ANC setup usually learns insecure cryptosystems even with a simple network designed to allow the One-Time Pad (OTP). It proposes CPA-ANC, a chosen-plaintext variant with a stronger Eve, under which the same network learns the OTP with high probability.
Key findings
- Without an adversary, all trials let Alice and Bob communicate without errors, but all learned algorithms were insecure; one example learned a permutation that did not use the key at all.
- Under the original ANC methodology, Alice and Bob could communicate and resist Eve, but the learned cryptosystems were usually not truly secure; an example used repeated key bits, and the paper concludes “very few” learned models were truly secure.
- Under the proposed CPA-ANC methodology, almost all successful models learned the OTP; only one model failed because Alice and Bob could not communicate. Examples show each plaintext bit XORed with a distinct key bit.
- The main claim is that an AI agent can learn a perfectly secure OTP without human knowledge, but the conditions are difficult and a very strong adversary is key to forcing a secure solution.
Methods
- Three neural-network agents (Alice, Bob, Eve) are used; Alice and Bob share a single simple CryptoNet that generalizes XOR via angle mappings \(f(b)=\arccos(1-2b)\) and \(f^{-1}(a)=(1-\cos(a))/2\). Eve is a ciphertext-only attacker in ANC and a classifier choosing between \(P_0\) and \(P_1\) in CPA-ANC.
- Training used random plaintexts and keys with minibatches of \(M=128\), key/plaintext lengths of 4, 8, and 16 bits, TensorFlow, and Adam with learning rate 0.0002. ANC alternated 5 minibatches for Alice/Bob and 10 for Eve, with \(\gamma=5\) and L2 regularization \(\alpha=0.1, 0.05, 0.01\); CPA-ANC alternated 3 minibatches for Alice/Bob and 60 for Eve, with \(\gamma=7\), L2 \(\alpha=0.1\) for 4/8 bits and \(\alpha=0.015\) for 16 bits, and Eve rules \(R=4n\).
- Stopping criteria were success if Bob’s decryption error was near zero and Eve’s attack was as bad as random guessing, or failure after 100,000 rounds. Learned models were tested with Algorithm 1 and judged secure if they learned the OTP.
Limitations
- The paper states it does not recommend using neural networks in real systems, that weak encryption is likely, and that it remains an open problem whether a neural network can learn a secure cryptosystem where a small key encrypts a very long message like a block or stream cipher.
- The paper states that CPA-ANC alone is not enough to guarantee security; designing a very strong adversary capable of breaking cryptosystems is key and remains a hard open problem.
- Evident from the text: experiments are limited to small keys (4, 8, 16 bits) and simple neural networks, security is reasoned about rather than formally proven, and in the ANC tests Eve was not retrained after Alice and Bob finished because the CryptoNet was simple enough to analyze manually.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures


![Figure 3 The proposed neural network (CryptoNet). The bits of the plaintext are represented by [p0, p1, …, pn−1]. The bits of the key are represented by [k0, k1, …, kn−1]. The function f (see Equation (9)) transforms the bits into angles [a0, a1, … a2n−1]. A fully connected layer combines the angles forming the variables [h0, h1, …, hn−1]. The function f⁻¹ (see Equation (10)) transforms the combined angles into continuous bits (real numbers in the interval [0,1]) [c0, c1, …, cn−1] representing the ciphertext.](https://figures.cochat.ai/paper-figures/2a46c00c07b04d2680544f040d8a0a50/learning-perfectly-secure-cryptography-to-protect-communications-with-figure-3.jpg)

![Figure 5 Eve’s neural network (CPA-CryptoNet). Eve receives as input two plaintexts P0 and P1. The bits of each plaintext are represented by [p0,0, …, p0,n−1] for P0 and [p1,0, …, p1,n−1] for P1. Eve also receives the ciphertext C represented by [c0, …, cn−1]. The function f (see Equation (9)) transforms the bits into angles [a0, a1, … a3n−1]. A fully connected layer combines the angles generating the hidden variables [h0⁰, h1⁰, …, hR−1⁰], where R is the number of rules. The function f⁻¹ (see Equation (10)) transforms the combined angles into continuous bits (real numbers in the interval [0,1]) [h0¹, h1¹, …, hR−1¹]. Another fully connected layer brings the hidden variables to logits through a softmax layer resulting in a probability π0 of C being a ciphertext of P0 and a probability π1 of C being a ciphertext of P1.](https://figures.cochat.ai/paper-figures/720f996e12be47268d68fb5f9fb5d9b9/learning-perfectly-secure-cryptography-to-protect-communications-with-figure-5.jpg)

Verification
Claims the paper makes
3 Valid · 0 Not valid · 3 Not checked · 33 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“As we will see in the experiments of Section 4, the problem with the approach proposed in the original ANC work [14] is that Eve’s job is too hard. It must decrypt a random message having access only to the ciphertext.”
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 3. Improvement to the ANC Methodology > 3.1. Chosen-Plaintext Attack Adversarial Neural Cryptography
What the sources say
- Supportedcites [14] Abadi, M., and Andersen, D.G. (arXiv, 2016). Learning to protect communications with adversarial neural cryptography, arXiv. source
Section 2.1 (p. 2) says Eve receives the ciphertext and attempts to recover the plaintext, without Bob’s secret key. Section 2.5 (p. 5) says the experiments use random plaintexts. Calling Eve’s task “too hard” is the citing authors’ own assessment, signaled by their reference to their forthcoming experiments, rather than a finding attributed to this source. (LLM, conf=0.90)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Using this neural network, we show that the ANC model is not good enough to generate secure cryptosystems even using a simple NN.”
- Also stated in: § 4.3. Training the network with Adversarial Neural Cryptography · boosted
“This demonstrates that the standard ANC technique proposed in [14] is not good enough because it usually does not lead to a secure cryptosystem even using a very simple NN designed to allow a perfectly secure solution.”
- Also stated in: § 6. Conclusions · boosted
“In addition, we have shown that the original ANC methodology is not good enough to achieve this goal.”
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 1. Introduction
What the sources say
- Partly supportedcites [14] Abadi, M., and Andersen, D.G. (arXiv, 2016). Learning to protect communications with adversarial neural cryptography, arXiv. sourceExperimental — accuracy not yet measured
The Introduction (p. 1) says the ANC method provides much weaker security guarantees than classical cryptography. It does not conclude that ANC cannot generate a secure cryptosystem, or test the particular simple neural network described in the citing sentence. That sentence presents the stronger result as the citing authors’ own finding. (LLM, conf=0.83)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 4.3. Training the network with Adversarial Neural Cryptography · boosted
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In addition, some of these cryptosystems are indeed Vigenère like ciphers, which can be broken.”
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 5. Comparison with Related Work
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The consequence is that, under this methodology, Alice and Bob learn to communicate with cryptosystems that are not truly secure.”
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 3. Improvement to the ANC Methodology > 3.1. Chosen-Plaintext Attack Adversarial Neural Cryptography
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Note that the key was not used in the learned algorithm of Equation (17).”
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 4. Results > 4.2. Training without an Adversary
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
19 more claims
Create a free account to see all 39 claims and every source verdict.
- Probably not a substantial claim. 7
- It may state two separate claims as one. 10
- The quote only partly states this claim. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“In particular, DL algorithms can perform fast and highly accurate evaluations on large datasets thanks to their ability to detect complex dependency structures and unpredictable patterns”
From A novel deep learning-based statistical randomness evaluation test methodology for cryptographic applications (2025) · cites this paper as Coutinho et al. 2018
Experimental — accuracy not yet measured
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.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 4. Results > 4.1. Method
“4-bit (n=4), 8-bit (n=8) and 16-bit (n=4).”
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 4. Results > 4.4. Learning the One-Time Pad
“16-bit (n=16) key length.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 3. Improvement to the ANC Methodology > 3.2. A Simple Neural Network Capable of Learning the One-Time Pad
“Note that the ciphertext is not composed of bits but by floating number between 0 and 1.”
§ Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography > 3. Improvement to the ANC Methodology > 3.2. A Simple Neural Network Capable of Learning the One-Time Pad
“P,K,C are vectors of n bits of input, key and output, respectively.”
Citations
Cites
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“Nowadays, several systems overcame human capabilities in human-like tasks as image recognition [1,2], speech recognition [3], driving cars [4] and playing intuitive games [5].”
BackgroundAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
TensorFlow: A system for large-scale machine learning
Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean, J., Devin, M., Ghemawat, S., Irving, G., Isard, M., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1605.08695 · Cited by 11,822
Communication Theory of Secrecy Systems
Shannon, C. E. (1949) · Bell System Technical Journal · doi:10.1002/j.1538-7305.1949.tb00928.x · Cited by 9,440 · cited 1×
“We designed this NN in a way that would be possible, but not necessary, to learn the One-Time Pad (OTP), which is well-known for being information-theoretically secure [15].”
MethodMastering the game of Go without human knowledge
Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., Bolton, A., et al. (2017) · Nature · doi:10.1038/nature24270 · Cited by 9,084 · cited 2×
“Nowadays, several systems overcame human capabilities in human-like tasks as image recognition [1,2], speech recognition [3], driving cars [4] and playing intuitive games [5].”
BackgroundResultSpeech recognition with deep recurrent neural networks
Graves, A., Mohamed, A.-R., & Hinton, G. (2013) · 2013 IEEE International Conference on Acoustics, Speech and Signal Processing · doi:10.1109/icassp.2013.6638947 · Cited by 8,893 · cited 1×
“Nowadays, several systems overcame human capabilities in human-like tasks as image recognition [1,2], speech recognition [3], driving cars [4] and playing intuitive games [5].”
BackgroundEnd to End Learning for Self-Driving Cars
Bojarski, M., Del Testa, D., Dworakowski, D., Firner, B., Flepp, B., Goyal, P., Jackel, L. D., Monfort, M., Muller, U., Zhang, J., et al. (2016) · arXiv · doi:10.48550/arxiv.1604.07316 · Cited by 3,198
Introduction to Modern Cryptography
Katz, J., & Lindell, Y. (2007) · doi:10.1201/9781420010756 · Cited by 1,238
A chaotic image encryption algorithm based on perceptron model
Wang, X.-Y., Yang, L., Liu, R., & Kadir, A. (2010) · Nonlinear Dynamics · doi:10.1007/s11071-010-9749-8 · Cited by 449
Learning to Protect Communications with Adversarial Neural Cryptography
Abadi, M., & Andersen, D. G. (2016) · arXiv · doi:10.48550/arxiv.1610.06918 · Cited by 182 · cited 10×
“A recent work defined the concept of ANC [14].”
BackgroundMethodResultInfluential
Research works printed but not linked: 5
- Simonyan, K., and Zisserman, A. (arXiv, 2014). Very deep convolutional networks for large-scale image recognition, arXiv.
- Li, C. (2005). Cryptanalyses of Some Multimedia Encryption Schemes. [Master’s Thesis, Zhejiang University].
- Desai, 2012, Using layer recurrent neural network to generate pseudo random number sequences, Int. J. Comput. Sci. Issues, 9, 324
- Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep Learning, MIT Press. Available online: http://www.deeplearningbook.org.
- Oliveira, L.B., Pereira, F.M.Q., Misoczki, R., Aranha, D.F., Borges, F., and Liu, J. (August, January 31). The computer for the 21st century: Security privacy challenges after 25 years. Proceedings of the 26th International Conference on Computer Communication and Networks (ICCCN), Vancouver, BC, Canada.
Cited by
Generative Adversarial Network Technologies and Applications in Computer Vision
Jin, L., Tan, F., & Jiang, S. (2020) · Computational Intelligence and Neuroscience · doi:10.1155/2020/1459107 · Cited by 118
Mehmood, A., Shafique, A., Alawida, M., & Khan, A. N. (2024) · IEEE Access · doi:10.1109/access.2024.3367232 · Cited by 118
“In [159], Coutinho et al. analyzed the security of cryptographic algorithms that are generated via adversarial neural cryptography (ANC).”
BackgroundResponsibility Gaps and Black Box Healthcare AI: Shared Responsibilization as a Solution
Lang, B. H., Nyholm, S. R., & Blumenthal‐Barby, J. (2023) · Digital Society · doi:10.1007/s44206-023-00073-z · Cited by 71
“…medical diagnosis (Rodriguez-Ruiz et al., 2019), predictive policing (McDaniel & Pease, 2021), recidivism prediction (Dressel & Farid, 2018), cryptography (Coutinho et al., 2018), autonomous weaponry for military and law enforcement (Wyatt, 2022), and self-driving cars (Joseph & Mondal, 2022).”
Neural Networks-Based Cryptography: A Survey
Meraouche, I., Dutta, S., Tan, H., & Sakurai, K. (2021) · IEEE Access · doi:10.1109/access.2021.3109635 · Cited by 52
Deep learning-based encryption for secure transmission digital images: A survey
Rohhila, S., & Singh, A. K. (2024) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2024.109236 · Cited by 57
Metadata
- Authors
- Coutinho, M., De Oliveira Albuquerque, R., Borges, F., García Villalba, L., & Kim, T.-H.
- Affiliations
- Universidade de Brasília (BR)
- Cybersecurity INCT Unit 6, Decision Technologies Laboratory—LATITUDE, Electrical Engineering Department (ENE), Technology College, University of Brasília (UnB), 70.910-900 Brasília-DF, Brazil
- Laboratório Nacional de Computação Científica (BR)
- Group of Analysis, Security and Systems (GASS), Department of Software Engineering and Artificial Intelligence (DISIA), Faculty of Computer Science and Engineering, Office 431, Universidad Complutense de Madrid (UCM), Calle Profesor José García Santesmases, 9, Ciudad Universitaria, 28040 Madrid, Spain
- Sungshin Women's University (KR)
- Venue
- Sensors
- Publisher
- MDPI AG
- Published
- 2018-04-24
- Volume
- 18
- Issue
- 5
- Pages
- 1306
- ISSN
- 1424-8220 (ISSN-L)
- PMC
- PMC5982701
- OpenAlex
- W2800109595
- Cited by
- 75
- References
- 17
- Citations by year
- 2026: 4 · 2025: 10 · 2024: 11 · 2023: 15 · 2022: 8 · 2021: 11 · 2020: 14 · 2019: 1 · 2018: 1
- FWCI
- 3.80
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Sensors (publishedVersion, cc-by)
- Europe PMC (PubMed Central) (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Library Open Repository (Universidad Complutense Madrid) (submittedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- European Commission (671672, H2020-ICT, H2020-ICT-2014, HORIZON2020, H2020, H2020-ICT-2014-2)
- Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (23038, 23038.007604/2014-69)
- Conselho Nacional de Desenvolvimento Científico e Tecnológico (2014‐2, 465741/2014-2)
- Fundação de Apoio à Pesquisa do Distrito Federal (0193.001366/2016, 0193.001365/2016)
- Forskningsrådet om Hälsa, Arbetsliv och Välfärd
- Horizon 2020 Framework Programme
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › Cryptography and Data Security
- Physical Sciences › Computer Science › Artificial Intelligence › Cryptographic Implementations and Security
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Keywords
- one-time pad, adversarial learning
- MeSH terms
- Algorithms, Artificial Intelligence, Communication
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-21, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-21 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography. https://app.cochat.ai/papers/10.3390/s18051306
- The paper
- Coutinho, M., De Oliveira Albuquerque, R., Borges, F., García Villalba, L., & Kim, T.-H. (2018). Learning Perfectly Secure Cryptography to Protect Communications with Adversarial Neural Cryptography. Sensors. https://doi.org/10.3390/s18051306
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.