Pith. sign in

REVIEW 5 major objections 5 minor 27 references

An Attack Method for Medical Insurance Claim Fraud Detection based on Generative Adversarial Network

T0 review · 5 major / 5 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 that a GAN trained on model outputs alone can generate fraudulent insurance claims that evade detection with 99% success.

desk verdict The 99% ASR claim is an artifact of circular evaluation; the dataset is mislabeled, the generator's output dimension doesn't match the features, and no held-out target is tested. read the letter →

arxiv 2506.19871 v1 pith:3I5664ZL submitted 2025-06-22 cs.CR cs.AI

Yining Pang , Chenghan Li This is my paper
classification cs.CRcs.AI
keywords insurancefrauddetectiongenerativeadversarialnetworkattackmedicalclaimsLSTMXGBoostblack-boxsuccessrate
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 claims that a generative adversarial network, trained with a surrogate-guided reinforcement-learning loop that only queries a fraud detector's outputs, can produce fraudulent medical insurance claims that the detector labels as legitimate. On a publicly available dataset of 1,000 claims with 38 features, the attack drives LSTM accuracy from 0.750 to 0.01 and XGBoost accuracy from 0.814 to 0.01, corresponding to a 99% attack success rate. The finding matters because it suggests that deployed insurance fraud detection systems are vulnerable to a practical black-box attack that requires no training data and no knowledge of the model's internals. The paper also reports that this GAN attack outperforms gradient-based baselines such as FGSM, BIM, and PGD, which need more access and achieve lower success.

What carries the argument

The machinery is a generative adversarial network whose generator maps random noise $z$ through five fully connected layers with 128, 256, 512, 64, and 12 units to produce synthetic samples $\tilde{x}=G(z)$. Training follows a surrogate-guided reinforcement-learning loop: a surrogate model $S$ classifies each generated sample, the binary prediction $\hat{y}=1[S(\tilde{x})>0.5]$ is compared against a target label $y_{\text{target}}$, and the reward feeds a temporal-difference update $z_{t+1}=z_t+\alpha\,\delta_t\,\gamma_t\,\mathcal{N}(0,I)$; the generator then minimizes a binary cross-entropy loss $L_G=\text{BCE}(S(\tilde{x}), y_{\text{target}})$ with Adam. The same trained generator is applied to fraud detection models that it did not see during training, which is what enables the claimed black-box transfer.

What would settle it

Reproduce the attack on the stated 1,000-sample, 38-feature insurance dataset using the described generator architecture with an output layer of 12 units, feed the generated samples directly into the LSTM and XGBoost fraud classifiers, and measure the attack success rate. If the classifiers reject the 12-dimensional inputs or the accuracy does not drop to about 0.01 once the dimension mismatch is handled, the 99% ASR claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a GAN can be trained to generate adversarial insurance claims that are classified as legitimate by fraud detectors, even when the attacker only has access to the model's output. The proposed method trains a generator through a reinforcement-learning scheme where a surrogate classifier supplies rewards; the generator is updated to produce samples that push the surrogate's prediction toward the legitimate class. Experiments on a public insurance dataset show the attack reduces the accuracy of an LSTM model from 0.750 to 0.01 and of an XGBoost model from 0.814 to 0.01, giving an attack success rate of 0.99 for both models. The authors interpret this as evidence that existing fraud detection systems, without adversarial defenses, can be effectively bypassed by an attacker who can query the model and does not know its training data or internal architecture.

Load-bearing premise

The load-bearing premise is that the generator's output, a 12-dimensional vector, can be treated as a complete 38-feature insurance claim that the fraud detectors will accept as input; if that mapping does not exist, the reported attack pipeline cannot run as written.

Editorial extensions

If this is right

  • Insurance fraud detectors that expose a prediction API without adversarial defenses can be bypassed by an attacker who only queries the model.
  • The attack transfers across different model families (LSTM and XGBoost), suggesting the vulnerability is architectural rather than model-specific.
  • The GAN attack outperforms gradient-based methods (FGSM, BIM, PGD) and random noise under the tested threat model, so defenses should be evaluated against this class of attack.
  • Deployed fraud detection pipelines should incorporate robustness measures such as adversarial training, input anomaly detection, or output-level filtering.
  • The demonstrated vulnerability implies that the financial losses from fraudulent claims could increase if attackers adopt GAN-based generation.

Reading between the lines

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

  • If the 99% figure holds, any fraud detector with a public query interface becomes exploitable without data or model access, which would push the security burden toward output-level defenses and rate-limiting rather than model secrecy.
  • The unstated mapping from the generator's 12-dimensional output to the dataset's 38 features is a likely gap; a repaired description would either select 12 features, pad to 38, or add a reconstruction layer, and the reported ASR would need to be re-verified under that mapping.
  • The black-box claim could be tested more forcefully by training the surrogate on a different model family from the target and measuring transfer; the paper does not report this, but it is a natural next experiment.
  • The same surrogate-guided GAN recipe should transfer to other tabular fraud domains, such as credit card or property claims, where the feature space is structured and classifiers are served behind an API.
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

5 major / 5 minor

Summary. The paper proposes a GAN-based adversarial attack against insurance claim fraud detectors. A generator is trained with a surrogate-guided reinforcement-learning reward (Eq. 2–4) to produce samples classified as non-fraudulent; experiments report that the attack reduces LSTM and XGBoost accuracy to 0.01, corresponding to a 99% attack success rate, and the authors argue this demonstrates black-box vulnerability. The paper also provides baseline comparisons with FGSM/BIM/PGD/random noise and SHAP analyses. The central claim—that an attacker with no knowledge of training data or model details can achieve 99% ASR—is not supported by the evidence as presented.

Significance. The topic is timely and the problem of adversarial robustness of insurance fraud detection is worth studying. The GAN-with-reinforcement-learning formulation is a reasonable design choice, and the paper includes a useful descriptive SHAP comparison of LSTM and XGBoost feature attributions. If the central claim were properly supported, this would be a useful contribution. However, the current evidence is insufficient: the 99% ASR is computed against models that appear to be the same ones used to train the generator, the generator's output dimension does not match the input space, and the dataset appears to be auto-insurance rather than medical claims. The paper provides no code or dataset link, so the experiments are not independently checkable.

major comments (5)
  1. [§2.2, Eq. (3)–(4), §3.4, Table 2] The central black-box claim is not established. In Section 2.2 the generator's reward is computed from the surrogate model S via Eq. (3) (thresholded prediction) and Eq. (4) (BCE against y_target), and Section 3.4 reports that the proposed attack reduces LSTM and XGBoost accuracy to 0.01 in Table 2. The paper never states that the LSTM and XGBoost models reported in Table 2 are held out from GAN training; if they are the same models used as S, the result only demonstrates that the generator was optimized on those models' outputs. Moreover, Section 3.4 says that gradient-based optimization is used for differentiable surrogates, which is stronger than the claimed 'accessing only the model output' scenario, and no transfer experiment to an unseen target model is reported. Without a held-out target, the abstract's claim of 99% ASR with no knowledge of training data or model details is unsupported.
  2. [§3.1 vs. §3.2] Section 3.1 describes a dataset with 1,000 samples and 38 features, while Section 3.2 states that the generator's final fully connected layer has 12 units. No projection, padding, feature-selection, or inverse-normalization step is described that would map the generator's 12-dimensional output to the 38-feature input space of the classifiers. As written, the attack pipeline cannot be executed, because the generated samples cannot be fed to the fraud detection models.
  3. [§3.1, Figures 3 and 5] The paper's title and abstract frame the contribution as a medical insurance claim fraud attack, but the features shown in Figures 3 and 5 (vehicle_claim, property_claim, injury_claim, insured_zip, policy_annual_premium, months_as_customer, capital-gains, capital-loss) are characteristic of an auto insurance claims dataset, not a medical claims dataset. The dataset is never named or cited, so the reader cannot resolve this mismatch. This undermines the relevance of the results to the stated application domain.
  4. [§3.1, Figure 4, Tables 1–2] The test split is inconsistent with the reported confusion matrices. Section 3.1 states a 75/5/20 split of 1,000 samples, which gives a 200-sample test set, but every confusion matrix in Figure 4 is labeled '250 Test Samples' and its entries sum to 250. In addition, the XGBoost baseline accuracy in Table 2 is 0.814, whereas Table 1 reports 0.825 for the same model. These numerical inconsistencies prevent the reader from verifying the attack success rate.
  5. [§3, Eq. (13), Table 2] The attack success rate is not defined in a way that can be reproduced. Eq. (13) defines ASR as '# batches deceiving stability / # batches sent', but 'deceiving stability' is undefined and no batch-level protocol is described. The claimed 99% ASR is inferred from the accuracy 0.01 in Table 2, yet the manuscript does not explain how the accuracy under attack is computed (e.g., whether generated samples replace all test samples, how labels are assigned, or how many samples are used).
minor comments (5)
  1. [Title] The rendered title contains 'F raud' with a spurious space; please fix the typo.
  2. [§2.3 and §3.2] The terms 'LightBoost' and 'LightGBM' are used interchangeably; standardize on one name.
  3. [§3.1] The dataset should be named and a URL or citation provided; currently it is described only as 'a publicly available insurance dataset'.
  4. [§3.1, Figure 3] Several panels in Figure 3 have '1e 5' in the y-axis label; the exponent formatting should be corrected.
  5. [§2.2] The latent variable is introduced as z ∈ R^{B×T×F} in Section 2.2, but the generator is later described as fully connected; the relationship between the tensor shape and the 12-unit output is not explained.

Circularity Check

1 steps flagged · score 6.0 of 10

The 99% ASR is measured on the models that supply the GAN's training reward, so the headline black-box attack result reduces to a fit, not an independent prediction.

  1. fitted input called prediction [Section 2.2 (Eqs. (2)-(4)) and Section 3.4 / Table 2]
    "inputs that maximize the classification error of a target or surrogate model S are learned to be synthesized by a generator G ... the reward signal is derived from the surrogate’s classification output ... LG =BCE(S(˜x),ytarget) (4) ... The results show an attack success rate (ASR) of 0.99 for both LSTM and XGBoost models"

    The generator is trained by minimizing LG = BCE(S(x~), ytarget), with the episodic reward also derived from the surrogate's classification output, and S is described as a target or surrogate model. Table 2 reports the attack's success on LSTM and XGBoost (accuracy 0.01, ASR 0.99) but never states that these evaluated models are distinct from S or that the generator was held out from them. If LSTM/XGBoost are S, then the reported ASR is the value of the training objective on the models that produced the reward: it measures optimization fit, not transfer to an unseen system. If they are distinct from S, no held-out transfer experiment is described.

full rationale

The derivation chain includes no imported uniqueness theorem and no self-citation: the GAN, the reinforcement-learning loss, and the classifiers are all described within this paper. The circularity lies in the evaluation protocol. Section 2.2 defines the generator's training objective as minimizing BCE(S(x~), y_target), where S is a target or surrogate model, and the episodic reward is the same surrogate output. Section 3.4 and Table 2 then report the 'Our' attack reducing accuracy to 0.01 on LSTM and XGBoost, with ASR 0.99. The paper never states that LSTM and XGBoost are distinct from S or were held out during generator training, so the reported 99% ASR is, by the paper's own equations, a measure of how well the generator minimized its reward on those exact classifiers; it is a fitted value rather than a prediction for an unseen black-box system. The generator's 12-unit output versus the dataset's 38 features and the statement that differentiable surrogates allow gradient-based optimization further undermine the black-box generalization claim, but those are correctness and specification issues rather than additional circular steps. Score 6 reflects that the central headline result reduces, in the as-written evaluation, to training against the evaluated models.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim depends on several hand-picked hyperparameters and unverified domain assumptions. The dataset is unnamed and appears mismatched to the paper's medical-insurance framing. The generator's output dimension is inconsistent with the feature count, and the RL update is stated without justification.

free parameters (4)
  • GAN layer sizes = 128, 256, 512, 64, 12
    Chosen by hand in Section 3.2 with no search or sensitivity analysis.
  • TD learning rate alpha and discount factor gamma
    Introduced in Eq. 2 in Section 2.2 without stated values or tuning procedure.
  • Baseline epsilon range = 0.05 to 0.50
    Selected by the authors for FGSM, BIM, PGD, and random noise attacks in Section 3.4.
  • Target label y_target = random or fixed
    Section 2.2 allows a randomly sampled or fixed target label; this choice is unconstrained and affects the attack success.
assumptions (4)
  • standard math Standard GAN minimax objective (Eq. 1) is assumed to be an effective generator of insurance claim-like inputs.
    Assumed in Section 2.1; no convergence analysis or adaptation to tabular data is given.
  • ad hoc to paper The temporal difference update z_{t+1} = z_t + alpha * delta_t * gamma_t * N(0,I) improves the generator's ability to fool the surrogate.
    Stated in Eq. 2 without derivation, ablation, or hyperparameter values; no evidence that this update is superior to standard GAN training.
  • domain assumption A surrogate model that outputs class probabilities provides a sufficient reward signal for the generator to produce attacks that transfer to the target.
    Section 2.2 relies on surrogate outputs for the RL reward; transfer to the target is never compared to a directly trained baseline.
  • domain assumption The unnamed 38-feature dataset represents medical insurance claim fraud.
    Section 3.1 describes the dataset only as 'publicly available'; feature names (vehicle_claim, property_claim, incident_hour) suggest auto insurance, and no dataset name or source is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Attack Method for Medical Insurance Claim Fraud Detection based on Generative Adversarial Network." pith.science (2026). https://pith.science/paper/3I5664ZL

@misc{pith2026250619871,
  author       = {Pith},
  title        = {Pith review of: An Attack Method for Medical Insurance Claim Fraud Detection based on Generative Adversarial Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3I5664ZL}},
  note         = {Machine review of arXiv:2506.19871}
}
read the original abstract

Insurance fraud detection represents a pivotal advancement in modern insurance service, providing intelligent and digitalized monitoring to enhance management and prevent fraud. It is crucial for ensuring the security and efficiency of insurance systems. Although AI and machine learning algorithms have demonstrated strong performance in detecting fraudulent claims, the absence of standardized defense mechanisms renders current systems vulnerable to emerging adversarial threats. In this paper, we propose a GAN-based approach to conduct adversarial attacks on fraud detection systems. Our results indicate that an attacker, without knowledge of the training data or internal model details, can generate fraudulent cases that are classified as legitimate with a 99\% attack success rate (ASR). By subtly modifying real insurance records and claims, adversaries can significantly increase the fraud risk, potentially bypassing compromised detection systems. These findings underscore the urgent need to enhance the robustness of insurance fraud detection models against adversarial manipulation, thereby ensuring the stability and reliability of different insurance systems.

Figures

Figures reproduced from arXiv: 2506.19871 by the authors.

Figure 1
Figure 1. Insurance Fraud Statistics[3] that integrates Convolutional Neural Networks (CNNs), Transformer architectures, and XGBoost to enhance detection accuracy and robustness. This framework not only incorporates domain-specific features, such as provider-patient interaction graphs and temporal patterns, but also enhances model interpretability through Shapley Additive exPlanations (SHAP) technology. Evaluated on two datas… view at source ↗
Figure 2
Figure 2. The Framework of GAN Attack Method 2.1 GAN Generative Adversarial Networks (GANs) are a class of deep learning models that have been widely used for unsupervised learning tasks. They consist of two neural networks: a generator and a discriminator, which are trained simultaneously through a zero-sum game framework.The generator G aims to produce synthetic data that is indistinguish￾able from real data. It takes a ran… view at source ↗
Figure 3
Figure 3. Distribution of characteristics of different categories [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Confusion matrices of Different Models The SHAP method can quantify the contribution of each feature to the model’s predictions and provide interpretability of the decision-making process. We used SHAP Gradient Explainer to interpret the LSTM model and SHAP Tree Explai…
Figure 5
Figure 5. Figure 5: SHAP values for LSTM and XGBoost assess the vulnerability of the model without any defense measures. For classical machine learning models with non-differentiable architectures, such as decision trees or ensemble methods, applying white-box adversarial attacks (e.g., F…
Figure 6
Figure 6. Figure 6: Model accuracy vs epsilon for LSTM. 0.0 0.1 0.2 0.3 0.4 0.5 Epsilon 0.0 0.2 0.4 0.6 0.8 1.0 Accuracy Accuracy vs Epsilon Random Noise [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Model accuracy vs epsilon for XGBoost. results show an attack success rate (ASR) of 0.99 for both LSTM and XGBoost models, highlighting the vulnerability of current insurance fraud detection models [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

27 extracted references · 22 canonical work pages

  1. [1]

    An intelligent machine learning approach for fraud detection in medical claim insurance: A comprehensive study.Scholars Journal of Engineering and Technology, 11(9):191–200, 2023

    Shashank Agarwal. An intelligent machine learning approach for fraud detection in medical claim insurance: A comprehensive study.Scholars Journal of Engineering and Technology, 11(9):191–200, 2023

    work page 2023

  2. [2]

    Implementation of a faith community nursing transition of care program in the usa: A propensity score matching analysis

    SangNam Ahn, Joonhyung Lee, Kathleen Munning, Katora P Campbell, Deborah Ziebarth, Lanet Owen, and Joel Jihwan Hwang. Implementation of a faith community nursing transition of care program in the usa: A propensity score matching analysis. Journal of Religion and Health, pages 1–18, 2025

    work page 2025

  3. [3]

    MdAlAmin,RushabhShah,HemanthTummala,andIndrajitRay. Utilizingblockchain andsmartcontractsforenhancedfraudpreventionandminimizationinhealthinsurance throughmulti-signatureclaimprocessing.In 2024InternationalConferenceonEmerging Trends in Networks and Computer Communications (ETNCC), pages 1–9. IEEE, 2024

    work page 2024

  4. [4]

    Albatul Albattah and Murad A Rassam. Detection of adversarial attacks against the hybrid convolutional long short-term memory deep learning technique for healthcare monitoring applications.Applied Sciences, 13(11):6807, 2023

    work page 2023

  5. [5]

    The mediating role of medical Title Suppressed Due to Excessive Length 13 service geographical availability between the healthcare service quality and the medical insurance

    MainNaserAlolayyan,AmalBaderAlnabelsi,Wa’edMohammadKhalafBaniSalameh, Najah Al-shanableh, Mazen Alzyoud, Mohammad Izzat Alhalalmeh, Mohammad Faleh Ahmmad Hunitie, Sulieman Ibraheem Mohammad, Anber Abraheem Shlash Mohammad, and Faraj Mazyed Faraj Aldaihani. The mediating role of medical Title Suppressed Due to Excessive Length 13 service geographical avail...

    work page 2025

  6. [6]

    Medicare fraud detection using machine learning methods

    Richard A Bauder and Taghi M Khoshgoftaar. Medicare fraud detection using machine learning methods. In2017 16th IEEE international conference on machine learning and applications (ICMLA), pages 858–865. IEEE, 2017

    work page 2017

  7. [7]

    Adversarial attack vulnerability of medical image analysis systems: Unexplored factors.Medical Image Analysis, 73:102141, 2021

    Gerda Bortsova, Cristina González-Gonzalo, Suzanne C Wetstein, Florian Dubost, Ioannis Katramados, Laurens Hogeweg, Bart Liefers, Bram van Ginneken, Josien PW Pluim, Mitko Veta, et al. Adversarial attack vulnerability of medical image analysis systems: Unexplored factors.Medical Image Analysis, 73:102141, 2021

    work page 2021

  8. [8]

    A survey on adversarial attacks and defences

    Anirban Chakraborty, Manaar Alam, Vishal Dey, Anupam Chattopadhyay, and Debdeep Mukhopadhyay. A survey on adversarial attacks and defences. CAAI Transactions on Intelligence Technology, 6(1):25–45, 2021

    work page 2021

Show all 27 references
  1. [9]

    Advancing fraud detection through deep learning: A comprehensive review

    Rakibul Hasan Chowdhury. Advancing fraud detection through deep learning: A comprehensive review. World Journal of Advanced Engineering Technology and Sciences, 12(2):606–613, 2024

    2024

  2. [10]

    Redefining insurance through technology: Achievements and perspectives in insurtech.Research in International Business and Finance, page 102301, 2024

    Simona Cosma and Giuseppe Rimo. Redefining insurance through technology: Achievements and perspectives in insurtech.Research in International Business and Finance, page 102301, 2024

    2024

  3. [11]

    Adversarial attacks on medical machine learning.Science, 363(6433):1287–1289, 2019

    Samuel G Finlayson, John D Bowers, Joichi Ito, Jonathan L Zittrain, Andrew L Beam, and Isaac S Kohane. Adversarial attacks on medical machine learning.Science, 363(6433):1287–1289, 2019

    2019

  4. [12]

    Explaining and harnessing adversarial examples.arXiv preprint arXiv:1412.6572, 2014

    Ian J Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and harnessing adversarial examples.arXiv preprint arXiv:1412.6572, 2014

    2014 arXiv

  5. [13]

    Big data fraud detection using multiple medicare data sources.Journal of Big Data, 5(1):1–21, 2018

    Matthew Herland, Taghi M Khoshgoftaar, and Richard A Bauder. Big data fraud detection using multiple medicare data sources.Journal of Big Data, 5(1):1–21, 2018

    2018

  6. [14]

    Comparingmedicareplanselectionamongbenefi- ciaries with and without a history of cancer.Health Affairs Scholar, 2(2):qxae014, 2024

    Shelley A Jazowski, Emma M Achola, Lauren H Nicholas, William A Wood, Christo- pherRFriese,andStacieBDusetzina. Comparingmedicareplanselectionamongbenefi- ciaries with and without a history of cancer.Health Affairs Scholar, 2(2):qxae014, 2024

    2024

  7. [15]

    Medicare fraud detection using neural networks

    Justin M Johnson and Taghi M Khoshgoftaar. Medicare fraud detection using neural networks. Journal of Big Data, 6(1):63, 2019

    2019

  8. [16]

    BP Kasaraneni. Machine learning models for fraud detection in health insurance claims: Techniques, applications, and real-world case studies.Journal of Machine Learning in Pharmaceutical Research, 4(1):110–147, 2024

    2024

  9. [17]

    Aulia-Absari Khalil, Ford Lumban Gaol, Boy Subirosa Sabarguna, and Harjanto Prabowo. A holistic platform for adverse drug reaction prevention and monitoring: leveraging service-oriented architecture for tailored healthcare solutions.Service Oriented Computing and Applications,...

    2025

  10. [18]

    Adversarial machine learning-industry perspectives

    Ram Shankar Siva Kumar, Magnus Nyström, John Lambert, Andrew Marshall, Mario Goertzel, Andi Comissoneru, Matt Swann, and Sharon Xia. Adversarial machine learning-industry perspectives. In2020 IEEE security and privacy workshops (SPW), pages 69–75. IEEE, 2020

    2020

  11. [19]

    Adversarial machine learning at scale.arXiv preprint arXiv:1611.01236, 2016

    Alexey Kurakin, Ian Goodfellow, and Samy Bengio. Adversarial machine learning at scale.arXiv preprint arXiv:1611.01236, 2016

    2016 arXiv

  12. [20]

    Future of generative adversarial networks (gan) for anomaly detection in network security: A review.Computers & Security, 139:103733, 2024

    Willone Lim, Kelvin Sheng Chek Yong, Bee Theng Lau, and Colin Choon Lin Tan. Future of generative adversarial networks (gan) for anomaly detection in network security: A review.Computers & Security, 139:103733, 2024

    2024

  13. [21]

    Towards deep learning models resistant to adversarial attacks.arXiv preprint arXiv:1706.06083, 2017

    Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu. Towards deep learning models resistant to adversarial attacks.arXiv preprint arXiv:1706.06083, 2017. 14 Y.Pang et al

    2017 arXiv

  14. [22]

    Adversarial robustness toolbox v1

    Maria-IrinaNicolae,MathieuSinn,MinhNgocTran,BeatBuesser,AmbrishRawat,Mar- tin Wistuba, Valentina Zantedeschi, Nathalie Baracaldo, Bryant Chen, Heiko Ludwig, et al. Adversarial robustness toolbox v1. 0.0.arXiv preprint arXiv:1807.01069, 2018

    2018 arXiv

  15. [23]

    Residual attention unet gan model for enhancing the intelligent agents in retinal image analysis

    Anuj Kumar Pandey, Satya Prakash Singh, and Chinmay Chakraborty. Residual attention unet gan model for enhancing the intelligent agents in retinal image analysis. Service Oriented Computing and Applications, pages 1–13, 2024

    2024

  16. [24]

    Syn-gan: A robust intrusion detection system using gan-based synthetic data for iot security

    Saifur Rahman, Shantanu Pal, Shubh Mittal, Tisha Chawla, and Chandan Karmakar. Syn-gan: A robust intrusion detection system using gan-based synthetic data for iot security. Internet of Things, 26:101212, 2024

    2024

  17. [25]

    Metaheuristic-based hyperparameter optimization for multi-disease detection and diagnosis in machine learning

    Jagandeep Singh, Jasminder Kaur Sandhu, and Yogesh Kumar. Metaheuristic-based hyperparameter optimization for multi-disease detection and diagnosis in machine learning. Service Oriented Computing and Applications, 18(2):163–182, 2024

    2024

  18. [26]

    Medicare fraud detection using graph analysis: A comparative study of machine learning and graph neural networks.IEEE Access, 11:88278–88294, 2023

    Yeeun Yoo, Jinho Shin, and Sunghyon Kyeong. Medicare fraud detection using graph analysis: A comparative study of machine learning and graph neural networks.IEEE Access, 11:88278–88294, 2023

    2023

  19. [27]

    Efficient adversarial training with transferable adversarial examples

    Haizhong Zheng, Ziqi Zhang, Juncheng Gu, Honglak Lee, and Atul Prakash. Efficient adversarial training with transferable adversarial examples. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1181–1190, 2020

    2020

Pith tools

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