Pith. sign in

REVIEW 4 major objections 6 minor 50 references

Fraud is Not Just Rarity: A Causal Prototype Attention Approach to Realistic Synthetic Oversampling

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 ↗ Review JSON

Pith's one-line read This paper argues that fraud oversampling works when the generator's latent space is shaped by a prototype attention classifier, not when the generator sees only frauds.

desk verdict A genuine but unproven architectural idea: the paper's central claim rests on an internal contradiction about what was trained and a 49-sample fraud test set, so it should go to peer review but not be trusted as a demonstration of the effect. read the letter →

arxiv 2507.14706 v2 pith:MKFW6HCF submitted 2025-07-19 cs.LG cs.AI

Claudio Giusti , Luca Guarnera , Mirko Casu , Sebastiano Battiato This is my paper · ORCID
classification cs.LGcs.AI
keywords creditcardfrauddetectionclassimbalancesyntheticoversamplingVAE-GANprototypeattentionclassifierlatentspaceshapinginterpretablemachinelearning
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 establish that the quality of synthetic fraud samples depends on the geometry of the latent space they come from, not just on the rarity of the fraud class. The authors argue that standard oversamplers, such as SMOTE interpolation and VAE-GANs trained only on frauds, fill in known fraud regions and can make classifiers overconfident because they never learn what separates fraud from normal transactions. Their proposal, CPAC, attaches a prototype-based attention classifier to the encoder of a VAE-GAN, so classification gradients shape the latent representation of both classes while the decoder generates new frauds. On the public credit-card fraud benchmark, they report F1 of 93.74% and recall of 92.85% for a gradient-boosted tree classifier trained with synthetic frauds from this pipeline, along with visibly tighter latent clusters. A sympathetic reader would care because it suggests a path from black-box augmentation to using the class boundary itself.

What carries the argument

The load-bearing object is CPAC, the Causal Prototype Attention Classifier: two learned prototype vectors $p_0, p_1 \in \mathbb{R}^d$, a per-feature attention mask $w=\mathrm{Att}(x)\in(0,1)^d$, and a learnable scale $\alpha>0$. It computes an attention-weighted squared distance $d_c(x)=\alpha\sum_i w_i (x_i - p_{c,i})^2$ to each prototype, treats the negative distances as logits, and outputs $\hat{y}=\mathrm{softmax}(-d_0, -d_1)$. In the joint VAE-GAN+CPAC training, the classifier's BCE loss, scale penalty, and prototype-anchoring penalty all backpropagate into the encoder, so the encoder is pushed to organize latent means around class prototypes while the decoder learns to generate frauds from the shaped space. This two-sided coupling, generative loss on reconstructions and classification loss on the same latent code, is what the paper claims turns sample generation into decision-aware synthetic oversampling. The paper uses "causal" loosely to mean that attention weights flag discriminative latent features; no causal graph or intervention is involved.

What would settle it

Train the VAE-GAN+CPAC pipeline twice, once with the generative loss restricted to fraud samples as the text describes and once on full-dataset batches as Algorithm 2 states; if downstream F1 and latent cluster separation are identical across the two runs, the CPAC head alone does not explain the reported improvement.

Watch

Extended reading notes

Core claim

The paper's central claim is that a classifier-guided generative pipeline can outperform pure sample augmentation. Concretely, the authors couple the encoder of a VAE-GAN with CPAC, a classifier that holds two learnable prototypes, one per class, weights each latent dimension by an attention mask, and scores a sample by its attention-weighted squared distance to the prototypes. Backpropagating the classification loss through the encoder while the generative decoder is trained produces a latent space with tight, separated clusters, and synthetic frauds drawn from this space improve downstream detection. The paper reports F1 of 93.74% and recall of 92.85% for a gradient-boosted tree classifier on the public credit-card fraud dataset, and it shows the same pipeline separating clusters on a second, independent synthetic credit-card dataset. It also argues that minority-only oversampling is the main source of the overconfidence problem, because fraud is defined relationally against normal transactions.

Load-bearing premise

The claim that the design differs from minority-only oversamplers depends on the generative losses being applied only to fraud samples while the classifier sees the full dataset, yet Algorithm 2 applies the VAE-GAN loss to every mini-batch of the full dataset; if the implementation follows the algorithm, the reported gain cannot be attributed to the CPAC head as described.

Editorial extensions

If this is right

  • Minority-only oversampling can be replaced by classifier-guided latent shaping, so generative oversamplers should be trained with access to both classes rather than on frauds alone.
  • Synthetic frauds generated from a CPAC-shaped latent space reduce overconfidence: downstream classifiers show more stable precision-recall behavior and do not need heavy augmentation to reach strong F1.
  • Prototype anchoring gives a built-in explanation for each prediction: a transaction is classified by its attention-weighted distance to the fraud prototype versus the normal prototype.
  • The same pipeline transfers to a second, independently generated credit-card dataset, suggesting the benefit is not specific to one benchmark.
  • A margin-aware ensemble classifier exploits the shaped latent space best, while simpler models such as logistic regression and nearest-neighbor classifiers gain less.

Reading between the lines

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

  • If the latent-geometry story holds, the number of synthetic samples matters less than where they land: the same downstream classifier should improve with few, well-placed synthetic frauds and degrade when the same count is drawn from an unshaped encoder.
  • The name 'causal' is not supported by the experiments; the attention weights are learned correlations. A direct test would intervene on a latent dimension, for example push a fraud sample toward the non-fraud prototype, and check whether the predicted fraud probability changes as the prototype-distance mechanism would predict.
  • The threshold-selection agent is presented as a separate component; an interesting extension would be to test whether its learned threshold transfers across datasets, since fraud priors differ.
  • A cleaner attribution experiment would compare generated samples from the CPAC-shaped encoder against samples from a vanilla VAE-GAN encoder with the same decoder and discriminator, holding sample count fixed, to isolate latent shaping from generation quality.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 Causal Prototype Attention Classifier (CPAC), an interpretable prototype-and-attention classification head, and couples it with a VAE-GAN encoder to shape a class-separable latent space for synthetic fraud oversampling. The authors evaluate the pipeline on the Kaggle credit card fraud dataset and an independent synthetic credit card dataset, comparing against SMOTE, a plain VAE-GAN oversampler, several MLP classifier heads, and recent baseline methods. They report that XGBoost trained on data augmented by the pre-oversampled VAE-GAN+CPAC pipeline achieves an F1-score of 93.74% and recall of 92.85%, with improved latent cluster separation. Ablation studies examine the role of the CPAC head, attention, prototypes, regularization penalties, and focal loss.

Significance. If the central claim were established, the work would provide a useful argument for shifting from minority-only generative oversampling toward classifier-guided latent shaping with interpretable prototype constraints. The paper has genuine strengths: the code is released, the ablation set is reasonably systematic (Section 5), an independent dataset is used for cross-validation (Section 4.5), and the Discussion candidly addresses overconfidence and generalization concerns (Section 6). However, the evidence as presented is undermined by an internal inconsistency between the textual description of the training protocol and the pseudocode, and by the statistical fragility of the headline performance claims. These issues are load-bearing for the paper's main assertion of superior performance, so the manuscript needs substantial revision before the contribution can be considered established.

major comments (4)
  1. [Section 3.6, Algorithm 2, Algorithm 3] The text in Section 3.6 states that 'the generative (VAE-GAN) component is optimized only on the minority class,' yet Algorithm 2 (lines 9-11) and Algorithm 3 (lines 10-12) compute L_VAE-GAN = L_recon + L_KL + L_GAN for every mini-batch drawn from the full training set, without any mask restricting the update to y=1. The classifier update in lines 13-15 of Algorithm 2 and lines 19-24 of Algorithm 3 also uses the same full-batch data. As written, the generative model is trained on both classes, not only on fraud. This is not a wording issue: the claimed distinction from minority-only oversamplers, and the attribution of the reported gains to CPAC's prototype anchoring and attention, depend on the generative loss being computed only on fraud samples. The ablation in Section 5.1 removes the CPAC head but retains the full-dataset VAE-GAN, so it does not isolate the contribution of CPAC under the claimed minority-only generative constraint. The authors must either correct the pseudocode to enforce a y=1 mask on the generative loss, or explicitly state and evaluate a protocol in which the generative component sees the full dataset; the current inconsistency prevents the reader from knowing which mechanism was actually tested.
  2. [Section 4.3.2, Table 7, Table 1] The headline comparison is statistically underdetermined. The test set contains only 49 fraud samples (Table 1), and the reported F1 advantage of the proposed pre-oversampled VAE-GAN+CPAC pipeline (93.74% for XGBoost at 100 synthetic samples, Table 7) over the best SMOTE result for XGBoost (93.15% at 75 samples, Table 4) is about 0.6 percentage points, i.e., less than the impact of a single correct or incorrect prediction on a 49-sample test set. No confidence intervals, bootstrap estimates, or repeated-seed runs are reported anywhere in the paper, so the observed differences cannot be distinguished from noise. The authors should report variability across seeds (or bootstraps) and an effect size or hypothesis test for the differences they claim as improvements.
  3. [Section 4.4, Tables 8 and 9] The state-of-the-art comparisons are not sufficiently controlled. Because the code for Shi et al. and Ahmed et al. is not available, the authors state that they 'reproduced their classification settings at the best of our abilities,' but no details are given about the hyperparameters, thresholds, training budgets, or number of runs for these reimplementations. Each reimplementation is then compared against the authors' XGBoost under different splits (75/25 for Shi et al., 80/20 for Ahmed et al. in Table 8), and no uncertainty is reported. Without a shared evaluation protocol and variance estimates, the claim that 'our XGBoost outperforms in every metric' is not established. The authors should provide full reproduction details, use identical preprocessing and evaluation splits for all methods, and report repeated-run statistics.
  4. [Section 5.1] The removal of the CPAC head in Section 5.1 is presented as evidence that 'the CPAC supervision is crucial for inducing clear separation,' but the configuration used for this ablation still trains the VAE-GAN on full batches of both classes, as in Algorithm 2 without the classifier head. This confounds the effect of the CPAC head with the effect of seeing majority-class data during generative training. A meaningful ablation would compare (a) full-data VAE-GAN with CPAC, (b) full-data VAE-GAN without CPAC, and (c) minority-only VAE-GAN with CPAC, so that the contribution of the supervised head can be separated from the influence of the training-data distribution.
minor comments (6)
  1. [Section 3.4] The term 'causal' in CPAC is acknowledged by the authors to be used 'loosely,' but this is potentially misleading; nothing in the model identifies a causal mechanism or estimates a causal effect. I suggest either removing the causal terminology or adding a concrete justification for why the attention weights should be interpreted causally.
  2. [Section 3.6.2, Equation 18] Equation 18 defines L_VAE-GAN as a sum of L_recon, L_KL, and L_GAN, but it is not stated whether L_GAN here is the generator's adversarial loss (Equation 6) or the discriminator loss (Equation 5); this ambiguity should be resolved in the text.
  3. [Section 4.2.3, Table 4] Several rows (notably KNN) report identical precision, recall, F1, and AUC across all oversampling levels, which is plausible for an instance-based method but deserves a brief explanation that the synthetic samples did not enter the nearest-neighbor geometry in a way that changed predictions.
  4. [Section 4.3.1] The pre-training oversampling step adds SMOTE samples before VAE-GAN+CPAC training, but it is not clear whether these SMOTE samples are also included in the training set for the downstream classifiers or only used to shape the latent space; this should be stated explicitly for reproducibility.
  5. [Abstract and Section 7] The abstract and conclusion claim 'superior performance' and 'improved F1-score, recall, and AUC' without qualification; given the statistical concerns in the major comments, I recommend softening these claims to reflect the uncertainty in the measurements.
  6. [Throughout] There are several typographical and formatting issues, such as inconsistent spacing in 'V AEs,' 'Y uhan,' and the use of 'state-of-the-art' versus 'SOTA'; a careful proofread would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the headline metrics are held-out test results, not derived quantities; the Section 3.6 vs Algorithm 2/3 contradiction is a correctness and attribution risk, not a circular reduction.

full rationale

This paper is an empirical method comparison rather than a derivation, and I find no step in which a predicted or reported quantity is equivalent to its inputs by construction. The central reported numbers (F1 93.74%, recall 92.85%) come from XGBoost evaluated on the held-out test set (Table 7), while CPAC's prototype anchoring, attention, and losses are training objectives; the test metric is not re-derived from those objectives. The only author self-citations ([14], [16]) appear in the introduction as background on deepfake detection and impostor bias and are not used to justify the central claim, so they are not load-bearing circularity. I do flag a load-bearing internal inconsistency: Section 3.6 states that 'the generative (VAE-GAN) component is optimized only on the minority class,' but Algorithm 2 (lines 9-11) and Algorithm 3 (lines 10-12) compute L_VAE-GAN = L_recon + L_KL + L_GAN on every mini-batch of the full training set with no class constraint. If the implementation follows the pseudocode, the generator is trained on both classes, so the reported improvement over minority-only oversamplers cannot be attributed to the CPAC head; the Section 5.1 ablation removes the head while retaining the full-data VAE-GAN, so it does not isolate CPAC under the claimed minority-only generative constraint. This is an attribution/soundness concern, not a circularity, because the reported metric does not reduce by definition to the method's inputs. Score 0 reflects the absence of circularity; the confound should be weighed as a correctness risk separately.

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

The paper introduces hyperparameters chosen by grid search or post-hoc selection (focal loss weights, anchor regularization, number of pre-SMOTE and generated samples). Its interpretability claim rests on the unverified assumption that attention weights are meaningful. The generative-training assumption is internally inconsistent, and the small test set weakens all metric comparisons.

free parameters (6)
  • alpha_FL = 0.95
    Focal loss class-balance weight, chosen by grid search to favor the minority class.
  • gamma = 2.0
    Focal loss focusing parameter, chosen by grid search to down-weight easy examples.
  • lambda_scale = 0.001
    Regularization weight for the scale penalty on the CPAC scale parameter, chosen by grid search.
  • lambda_anchor = 0.01
    Regularization weight for the prototype anchoring penalty, chosen by grid search.
  • SMOTE_pre_samples = 75
    Number of SMOTE samples added before VAE-GAN+CPAC training, selected based on prior experiments (Section 4.3.1).
  • generated_samples = 100 (for final comparison)
    Number of synthetic fraud samples added to the training set; the final headline result uses 100, chosen from {50, 75, 100} based on performance.
assumptions (5)
  • domain assumption The Kaggle credit card dataset is representative of real-world fraud.
    The evaluation and conclusions are based on this public dataset; the paper does not test on a real private fraud stream.
  • domain assumption The independent synthetic dataset (Kartik Shenoy) is a valid proxy for real transaction patterns.
    This dataset is generated by a simulator (Sparkov), so generalization claims to real fraud are limited.
  • ad hoc to paper Attention weights in CPAC indicate feature relevance for classification.
    The paper calls this 'causal' but cites works showing attention is not explanation (refs 36, 37); the assumption is not independently validated.
  • ad hoc to paper The generative loss is computed exclusively on minority-class samples while the encoder receives full-data supervision.
    Stated in Section 3.6 but contradicted by Algorithm 2, where the VAE-GAN loss is computed on every mini-batch from the full dataset.
  • domain assumption The test set's 49 fraud samples are sufficient for reliable metric estimation.
    With 49 positives, a single prediction changes recall by about 2 percentage points; the paper reports point estimates without error bars.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fraud is Not Just Rarity: A Causal Prototype Attention Approach to Realistic Synthetic Oversampling." pith.science (2026). https://pith.science/paper/MKFW6HCF

@misc{pith2026250714706,
  author       = {Pith},
  title        = {Pith review of: Fraud is Not Just Rarity: A Causal Prototype Attention Approach to Realistic Synthetic Oversampling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MKFW6HCF}},
  note         = {Machine review of arXiv:2507.14706}
}
read the original abstract

Detecting fraudulent credit card transactions remains a significant challenge, due to the extreme class imbalance in real-world data and the often subtle patterns that separate fraud from legitimate activity. Existing research commonly attempts to address this by generating synthetic samples for the minority class using approaches such as GANs, VAEs (Variational Autoencoders), or hybrid generative models. However, these techniques, particularly when applied only to minority-class data, tend to result in overconfident classifiers and poor latent cluster separation, ultimately limiting real-world detection performance. In this study, we propose the Causal Prototype Attention Classifier (CPAC), an interpretable architecture that promotes class-aware clustering and improved latent space structure through prototype-based attention mechanisms and we couple it with the encoder of a Variational Autoencoder-Generative Adversarial Network (VAE-GAN) in order to achieve improved latent cluster separation moving beyond post-hoc sample augmentation. We compared CPAC-augmented models to traditional oversamplers, such as SMOTE, as well as to state-of-the-art generative models, both with and without CPAC-based latent classifiers. Our results show that classifier-guided latent shaping with CPAC delivers superior performance, achieving an F1-score of 93.74% and recall of 92.85%, along with improved latent cluster separation. Further ablation studies and visualizations provide deeper insight into the benefits and limitations of classifier-driven representation learning for fraud detection. The codebase for this work can be found at the following link: https://github.com/claudiunderthehood/VAEGAN-CPAC.git.

Figures

Figures reproduced from arXiv: 2507.14706 by the authors.

Figure 1
Figure 1. PCA plots comparing frauds distribution before and after SMOTE oversampling. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Diagram portraying the structure of the VAE-GAN. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. PCA plots comparing frauds distribution before and after VAE-GAN oversampling. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Architecture of the CPAC model. Each input is compared to class prototypes using an attention-weighted [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Architectures of the three MLP heads used for classification on the 2-dimensional latent space. Hidden units [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: PCA plots showing cluster separation in the latent space for different MLP heads. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Latent space visualizations for the Encoder with CPAC head: (a) PCA, (b) 3D overlap. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Latent space visualizations with CPAC head and SMOTE pre-oversampling. (a) PCA, (b) 3D overlap. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: 2D PCA plot of the cluster representation of the Encoder with CPAC head trained with SMOTE pre [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Latent space visualizations for the Encoder with CPAC head obtained with the Synthetic Dataset. The two [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: 2D PCA plot of the cluster representation of the Encoder without CPAC. [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]
Figure 12
Figure 12. Figure 12: 2D PCA plot of the cluster representation of the Encoder of VAEGAN+CPAC without the attention [PITH_FULL_IMAGE:figures/full_fig_p021_12.png]
Figure 13
Figure 13. Figure 13: 2D PCA plot of the cluster representation of the Encoder of VAEGAN+CPAC without prototypes. [PITH_FULL_IMAGE:figures/full_fig_p022_13.png]
Figure 14
Figure 14. Figure 14: 2D PCA plot of the cluster representation of the Encoder of VAEGAN+CPAC without scale and anchor [PITH_FULL_IMAGE:figures/full_fig_p022_14.png]
Figure 15
Figure 15. Figure 15: Latent space visualizations for VAE-GAN+CPAC using Focal Loss: (a) 2D PCA plot; (b) 3D overlap [PITH_FULL_IMAGE:figures/full_fig_p023_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 48 canonical work pages

  1. [1]

    Outside the Closed World: On Using Machine Learning for Network Intrusion Detection

    Robin Sommer and Vern Paxson. Outside the Closed World: On Using Machine Learning for Network Intrusion Detection. In2010 IEEE Symposium on Security and Privacy, 2010

    work page 2010

  2. [2]

    García, M

    S. García, M. Grill, J. Stiborek, and A. Zunino. An empirical comparison of botnet detection methods.Computers & Security, 45:100–123, 2014

    work page 2014

  3. [3]

    Bhuyan, D

    Monowar H. Bhuyan, D. K. Bhattacharyya, and J. K. Kalita. Network Anomaly Detection: Methods, Systems and Tools.IEEE Communications Surveys & Tutorials, 16(1):303–336, 2014

    work page 2014

  4. [4]

    The base-rate fallacy and the difficulty of intrusion detection.ACM Trans

    Stefan Axelsson. The base-rate fallacy and the difficulty of intrusion detection.ACM Trans. Inf. Syst. Secur., 3(3):186–205, 2000

    work page 2000

  5. [5]

    Anomaly detection: A survey.ACM Comput

    Varun Chandola, Arindam Banerjee, and Vipin Kumar. Anomaly detection: A survey.ACM Comput. Surv., 41(3), 2009

    work page 2009

  6. [6]

    Bridges, Tarrah R

    Robert A. Bridges, Tarrah R. Glass-Vanderlan, Michael D. Iannacone, Maria S. Vincent, and Qian (Guenevere) Chen. A Survey of Intrusion Detection Systems Leveraging Host Data.ACM Comput. Surv., 52(6), 2019

    work page 2019

  7. [7]

    Examining the costs and causes of cyber incidents.Journal of Cybersecurity, 2(2):121–135, 2016

    Sasha Romanosky. Examining the costs and causes of cyber incidents.Journal of Cybersecurity, 2(2):121–135, 2016

    work page 2016

  8. [8]

    Characterizing fraud and its ramifications in affiliate marketing networks.Journal of Cybersecurity, 2(1):71–81, 2016

    Peter Snyder and Chris Kanich. Characterizing fraud and its ramifications in affiliate marketing networks.Journal of Cybersecurity, 2(1):71–81, 2016

    work page 2016

Show all 50 references
  1. [9]

    International comparison of bank fraud reimbursement: customer perceptions and contractual terms.Journal of Cybersecurity, 3(2):109–125, 2018

    Ingolf Becker, Alice Hutchings, Ruba Abu-Salma, Ross Anderson, Nicholas Bohm, Steven J Murdoch, M Angela Sasse, and Gianluca Stringhini. International comparison of bank fraud reimbursement: customer perceptions and contractual terms.Journal of Cybersecurity, 3(2):109–125, 2018

    2018

  2. [10]

    Intelligent financial fraud detection: A comprehensive review.Computers & Security, 57:47–66, 2016

    Jarrod West and Maumita Bhattacharya. Intelligent financial fraud detection: A comprehensive review.Computers & Security, 57:47–66, 2016

    2016

  3. [11]

    Falcone Sampaio

    Michael Edward Edge and Pedro R. Falcone Sampaio. A survey of signature based methods for financial fraud detection.Computers & Security, 28(6):381–394, 2009

    2009

  4. [12]

    Reducing false positives in bank anti-fraud systems based on rule induction in distributed tree-based models.Computers & Security, 120:102786, 2022

    Ivan V orobyev and Anna Krivitskaya. Reducing false positives in bank anti-fraud systems based on rule induction in distributed tree-based models.Computers & Security, 120:102786, 2022. 25 Fraud is Not Just RarityA PREPRINT

    2022

  5. [13]

    Detection of AI deepfake and fraud in online payments using gan-based models.arXiv preprint arXiv:2501.07033, 2025

    Zong Ke, Shicheng Zhou, Yining Zhou, Chia Hong Chang, and Rong Zhang. Detection of AI deepfake and fraud in online payments using gan-based models.arXiv preprint arXiv:2501.07033, 2025

    2025

  6. [14]

    Deepfake media forensics: Status and future challenges.Journal of Imaging, 11(3):73, 2025

    Irene Amerini, Mauro Barni, Sebastiano Battiato, Paolo Bestagini, Giulia Boato, Vittoria Bruni, Roberto Caldelli, Francesco De Natale, Rocco De Nicola, Luca Guarnera, et al. Deepfake media forensics: Status and future challenges.Journal of Imaging, 11(3):73, 2025

    2025

  7. [15]

    Banking in the Age of Deepfakes: Evaluating Perceptions of Deepfake Fraud Risks

    Ambu Sharma and Rajesh Tiwari. Banking in the Age of Deepfakes: Evaluating Perceptions of Deepfake Fraud Risks. InNavigating the World of Deepfake Technology, pages 454–469. IGI Global Scientific Publishing, 2024

    2024

  8. [16]

    GenAI mirage: The impostor bias and the deepfake detection challenge in the era of artificial illusions.Forensic Science International: Digital Investigation, 50:301795, 2024

    Mirko Casu, Luca Guarnera, Pasquale Caponnetto, and Sebastiano Battiato. GenAI mirage: The impostor bias and the deepfake detection challenge in the era of artificial illusions.Forensic Science International: Digital Investigation, 50:301795, 2024

    2024

  9. [17]

    Wiley, 3rd edition, 2013

    David W Hosmer, Stanley Lemeshow, and Rodney X Sturdivant.Applied Logistic Regression. Wiley, 3rd edition, 2013

    2013

  10. [18]

    Random Forests.Machine Learning, 45(1):5–32, 2001

    Leo Breiman. Random Forests.Machine Learning, 45(1):5–32, 2001

    2001

  11. [19]

    Xgboost: A scalable tree boosting system

    Tianqi Chen and Carlos Guestrin. Xgboost: A scalable tree boosting system. pages 785–794, 2016

    2016

  12. [20]

    SMOTE: synthetic minority over-sampling technique.Journal of artificial intelligence research, 16:321–357, 2002

    Nitesh V Chawla, Kevin W Bowyer, Lawrence O Hall, and W Philip Kegelmeyer. SMOTE: synthetic minority over-sampling technique.Journal of artificial intelligence research, 16:321–357, 2002

    2002

  13. [21]

    Auto-Encoding Variational Bayes.arXiv preprint arXiv:1312.6114, 2014

    Diederik P Kingma and Max Welling. Auto-Encoding Variational Bayes.arXiv preprint arXiv:1312.6114, 2014

    2014 arXiv

  14. [22]

    Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative Adversarial Nets. In Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, and K.Q. Weinberger, editors,Advances in Neural Information Pr...

    2014

  15. [23]

    Denoising Diffusion Probabilistic Models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising Diffusion Probabilistic Models. In H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin, editors,Advances in Neural Information Processing Systems, volume 33, pages 6840–6851. Curran Associates, Inc., 2020

    2020

  16. [24]

    Deep Unsupervised Learning using Nonequilibrium Thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep Unsupervised Learning using Nonequilibrium Thermodynamics. In Francis Bach and David Blei, editors,Proceedings of the 32nd International Conference on Machine Learning, volume 37 ofProceedings of ...

    2015

  17. [25]

    Deep learning for credit card fraud detection.Journal of Financial Data Science, 1(1):1–12, 2018

    Yixuan Li, Feng Xie, Jian Wang, and Yajie Cai. Deep learning for credit card fraud detection.Journal of Financial Data Science, 1(1):1–12, 2018

    2018

  18. [26]

    Credit Card Fraud Detection Dataset, 2015

    Andrea Dal Pozzolo, Giacomo Boracchi, Olivier Caelen, Cesare Alippi, and Gianluca Bontempi. Credit Card Fraud Detection Dataset, 2015. Accessed via Kaggle. Published by Worldline and the Machine Learning Group of ULB

    2015

  19. [27]

    Explainable AI: A Review of Machine Learning Interpretability Methods.Entropy, 23(1), 2021

    Pantelis Linardatos, Vasilis Papastefanopoulos, and Sotiris Kotsiantis. Explainable AI: A Review of Machine Learning Interpretability Methods.Entropy, 23(1), 2021

    2021

  20. [28]

    Chawla, Kevin W

    Nitesh V . Chawla, Kevin W. Bowyer, Lawrence O. Hall, and W. Philip Kegelmeyer. SMOTE: Synthetic minority over-sampling technique.Journal of Artificial Intelligence Research, 16:321–357, 2002

    2002

  21. [29]

    Prati, Bartosz Krawczyk, and Francisco Herrera

    Alberto Fernández, Salvador García, Mikel Galar, Ricardo C. Prati, Bartosz Krawczyk, and Francisco Herrera. SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-Year Anniversary.Journal of Artificial Intelligence Research, 61:863–905, 2018

    2018

  22. [30]

    Application of Deep Generative Models for Anomaly Detection in Complex Financial Transactions.arXiv preprint arXiv:2504.15491, 2025

    Tingting Tang, Jing Yao, Yuhan Wang, Qiang Sha, Huan Feng, and Zhi Xu. Application of Deep Generative Models for Anomaly Detection in Complex Financial Transactions.arXiv preprint arXiv:2504.15491, 2025

    2025 arXiv

  23. [31]

    A Survey on GAN Techniques for Data Augmentation to Address the Imbalanced Data Issues in Credit Card Fraud Detection.Machine Learning and Knowledge Extraction, 5(1):304–329, 2023

    Emilija Strelcenia and Simant Prakoonwit. A Survey on GAN Techniques for Data Augmentation to Address the Imbalanced Data Issues in Credit Card Fraud Detection.Machine Learning and Knowledge Extraction, 5(1):304–329, 2023

    2023

  24. [32]

    This Looks Like That: Deep Learning for Interpretable Image Recognition

    Chaofan Chen, Oscar Li, Daniel Tao, Alina Barnett, Cynthia Rudin, and Jonathan K Su. This Looks Like That: Deep Learning for Interpretable Image Recognition. InAdvances in Neural Information Processing Systems, editor = H. Wallach and H. Larochelle and A. Beygelzimer and F . d...

    2019

  25. [33]

    This Looks Like That, Because

    Meike Nauta, Annemarie Jutte, Jesper Provoost, and Christin Seifert. This Looks Like That, Because ... Explaining Prototypes for Interpretable Image Recognition. InMachine Learning and Principles and Practice of Knowledge Discovery in Databases, pages 441–456, Cham, 2021. Spri...

    2021

  26. [34]

    Interpretable and Steerable Sequence Learning via Prototypes

    Ming Yao, Pengxu Xu, Huamin Qu, and Liu Ren. Interpretable and Steerable Sequence Learning via Prototypes. arXiv preprint arXiv:1907.09728, 2019

    1907 arXiv

  27. [35]

    This Looks Like That

    Andreas Hoffmann, Carlo Fanconi, Raphael Rade, and Jan Kohler. This Looks Like That... Does It? Shortcomings of Latent Space Prototype Interpretability in Deep Networks.arXiv preprint arXiv:2105.02968, 2021

    2021 arXiv

  28. [36]

    Sarthak Jain and Byron C. Wallace. Attention is not Explanation. InProceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 3543–3556, Minneapolis, Mi...

    2019

  29. [37]

    Attention is not not Explanation

    Sarah Wiegreffe and Yuval Pinter. Attention is not not Explanation. InProceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 11–20, Hong Kong, China,...

    2019

  30. [38]

    Why Should I Trust You?

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. “Why Should I Trust You?": Explaining the Predictions of Any Classifier. InProceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’16, page 1135–1144, New York, NY , USA, 2...

    2016

  31. [39]

    A Unified Approach to Interpreting Model Predictions

    Scott M Lundberg and Su-In Lee. A Unified Approach to Interpreting Model Predictions. InAdvances in Neural Information Processing Systems, volume 30. Curran Associates, Inc., 2017

    2017

  32. [40]

    An ensemble-based credit card fraud detection algorithm using an efficient voting strategy.The Computer Journal, 65(8):1998–2015, 05 2021

    Morteza Rakhshaninejad, Mohammad Fathian, Babak Amiri, and Navid Yazdanjue. An ensemble-based credit card fraud detection algorithm using an efficient voting strategy.The Computer Journal, 65(8):1998–2015, 05 2021

    1998

  33. [41]

    Wang and Y

    D. Wang and Y . Yao. Unrolled GAN-Based Oversampling of Credit Card Dataset for Fraud Detection. In2022 2nd International Conference on Artificial Intelligence and Computer Applications (ICAICA), pages 722–727, Dalian, China, 2022. IEEE

    2022

  34. [42]

    Credit Card Fraud Detection Based on Improved Variational Autoencoder Generative Adversarial Network.IEEE Access, 11:84545–84556, 2023

    Weijun Ding, Qian Kang, and Yuling Feng. Credit Card Fraud Detection Based on Improved Variational Autoencoder Generative Adversarial Network.IEEE Access, 11:84545–84556, 2023

    2023

  35. [43]

    An attention-based balanced variational autoencoder method for credit card fraud detection.Applied Soft Computing, 177:113190, 2025

    Si Shi, Wuman Luo, and Giovanni Pau. An attention-based balanced variational autoencoder method for credit card fraud detection.Applied Soft Computing, 177:113190, 2025

    2025

  36. [44]

    A credit card fraud detection approach based on ensemble machine learning classifier with hybrid data sampling.Machine Learning with Applications, 20:100675, 2025

    Khanda Hassan Ahmed, Stefan Axelsson, Yuhong Li, and Ali Makki Sagheer. A credit card fraud detection approach based on ensemble machine learning classifier with hybrid data sampling.Machine Learning with Applications, 20:100675, 2025

    2025

  37. [45]

    Alejo, J

    R. Alejo, J. M. Sotoca, R. M. Valdovinos, and P. Toribio. Edited Nearest Neighbor Rule for Improving Neural Networks Classifications. In Liqing Zhang, Bao-Liang Lu, and James Kwok, editors,Advances in Neural Networks - ISNN 2010, pages 303–310. Springer Berlin Heidelberg, 2010

    2010

  38. [46]

    Indrajani Sutedja, Yaya Heryadi, Lili Ayu Wulandhari, and Bachtiar Abbas. Imbalanced data classification using auxiliary classifier generative adversarial networks.International Journal of Advanced Trends in Computer Science and Engineering, 9(2):1068–1075, 2020

    2020

  39. [47]

    Conditional wasserstein gan-based oversampling of tabular data for imbalanced learning.Expert Systems with applications, 174:114582, 2021

    Justin Engelmann and Stefan Lessmann. Conditional wasserstein gan-based oversampling of tabular data for imbalanced learning.Expert Systems with applications, 174:114582, 2021

    2021

  40. [48]

    Focal Loss for Dense Object Detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. Focal Loss for Dense Object Detection. InProceedings of the IEEE International Conference on Computer Vision (ICCV), ICCV ’17, pages 2980–2988, New York, NY , USA, 2017. IEEE

    2017

  41. [49]

    Credit Card Transactions Fraud Detection Dataset, 2020

    Kartik Shenoy. Credit Card Transactions Fraud Detection Dataset, 2020. Accessed via Kaggle. Published by Kartik Shenoy

    2020

  42. [50]

    Sparkov Data Generator Tool for Generating Credit Card Transaction Data, 2022

    Brandon Harris. Sparkov Data Generator Tool for Generating Credit Card Transaction Data, 2022. Accessed via Github. Published by Brandon Harris. 27

    2022

Pith tools

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