Adversarial Robustness for Tabular Data through Cost and Utility Awareness

Kireev, Klim; Kulynych, Bogdan; Troncoso, Carmela

doi:10.14722/ndss.2023.24924

Cited by 6 publications

(3 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our proposed method achieves a TPR of 91.7% for fraud and 100% for the non-fraud class as shown in Table 6. Another study [51] used the BAF dataset to check the adversarial robustness of text and image data and achieved an accuracy of 67.7% using gradient-boosted Stamps while our study achieved an accuracy of 95.8% for both classes.…”

Section: Comparison With State-of-the-art Methods In New Bank Account...mentioning

confidence: 74%

Modeling the Dynamic Behaviors of Bank Account Fraudsters Using Combined Simultaneous Game Theory with Neural Networks

USMAN,

Abdullahi,

Ran

et al. 2024

Preprint

View full text Add to dashboard Cite

New bank account fraud is a significant problem causing financial losses in banking and finance. Existing statistical and machine-learning methods were used to detect fraud thereby preventing financial losses. However, most studies do not consider the dynamic behavior of fraudsters and often produce a high False Positive Rate (FPR). This study proposes the detection of new bank account fraud in the context of simultaneous game theory (SGT) with Neural Networks, the SGT involves two players, a fraudster, and bank officials attacking each other through Bayesian probability in a zero-sum. The influence of outliers within the SGT was tackled by adding a context feature for effective simulation of the dynamic behavior of fraudsters. The Neural Networks layer uses the simulated features for fraud context learning. The study is validated using Bank Account Fraud (BAF) Dataset on different machine-learning models. The Radial Basis Function Networks achieved FPR of 0.0% and 8.3% for fraud and non-fraud classes, respectively, while achieving True Positive Rate (TPR) of 91.7% and 100.0% for fraud and non-fraud classes, respectively. An improved Radial Basis Function Networks detect fraud by revealing fraudulent patterns and dynamic behaviors in higher dimensional data. The findings will enhance fraud detection and reduce customer attrition.

show abstract

Section: Comparison With State-of-the-art Methods In New Bank Account...mentioning

confidence: 74%

Modeling the Dynamic Behaviors of Bank Account Fraudsters Using Combined Simultaneous Game Theory with Neural Networks

USMAN,

Abdullahi,

Ran

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…A point of note is that most of the published papers refer to attacks and defenses on image data [15]. Only in recent years, the interest in attacks and defense on tabular data processing models has increased [16].…”

Section: A Advesarial Machine Learning (Aml)mentioning

confidence: 99%

Towards automated detection of adversarial attacks on tabular data

Biczyk,

Wawrowski

2023

Annals of Computer Science and Information Systems

View full text Add to dashboard Cite

The paper presents a novel approach to investigating adversarial attacks on machine learning classification models operating on tabular data. The employed method involves using diagnostic parameters calculated on an approximated representation of a model under attack and analyzing differences in these diagnostic parameters over time. The hypothesis researched by the authors is that adversarial attack techniques, even if attempting a low-profile modification of input data, influence those diagnostic attributes in a statistically significant way. Thus, changes in diagnostic attributes can be used for detecting attack events. Three attack approaches on real-world datasets were investigated. The experiments confirm the approach as a promising technique to be further developed for detecting adversarial attacks.

show abstract

“…Most AA detection techniques focus on image processing models [10], with tabular data techniques receiving attention only in recent years [11].…”

mentioning

confidence: 99%

Detecting and Isolating Adversarial Attacks Using Characteristics of the Surrogate Model Framework

Biczyk,

Wawrowski

2023

Applied Sciences

View full text Add to dashboard Cite

The paper introduces a novel framework for detecting adversarial attacks on machine learning models that classify tabular data. Its purpose is to provide a robust method for the monitoring and continuous auditing of machine learning models for the purpose of detecting malicious data alterations. The core of the framework is based on building machine learning classifiers for the detection of attacks and its type that operate on diagnostic attributes. These diagnostic attributes are obtained not from the original model, but from the surrogate model that has been created by observation of the original model inputs and outputs. The paper presents building blocks for the framework and tests its power for the detection and isolation of attacks in selected scenarios utilizing known attacks and public machine learning data sets. The obtained results pave the road for further experiments and the goal of developing classifiers that can be integrated into real-world scenarios, bolstering the robustness of machine learning applications.

show abstract

Adversarial Robustness for Tabular Data through Cost and Utility Awareness

Cited by 6 publications

References 61 publications

Modeling the Dynamic Behaviors of Bank Account Fraudsters Using Combined Simultaneous Game Theory with Neural Networks

Modeling the Dynamic Behaviors of Bank Account Fraudsters Using Combined Simultaneous Game Theory with Neural Networks

Towards automated detection of adversarial attacks on tabular data

Detecting and Isolating Adversarial Attacks Using Characteristics of the Surrogate Model Framework

Contact Info

Product

Resources

About