Yong Xia

Title: Yong Xia and Adversarially Hardened Machine Learning

Introduction

Yong Xia is an inventor based in Guangzhou, China. His patent work addresses ways to strengthen machine-learning systems against adversarial attacks and improve fraud detection.

Latest Patents

Xia is associated with 1 patent, titled “Systems and Methods for Machine Learning Using Adversarial Learning and Data Transformation.” The invention describes a system and method for hardening machine-learning models against adversarial attacks, including challenges caused by imbalanced data and sophisticated attempts to evade detection.

The method preprocesses data for logistic regression and uses a Negative-Calibrated Generative Adversarial Network (NCGAN) to generate synthetic fraudulent transactions and balance the dataset. It models adversarial behavior through a Stackelberg game framework, then trains a logistic regression model with a sum of top-K loss function. By emphasizing transactions with the greatest loss, the approach aims to improve the detection of fraudulent activity.

Career Highlights

Xia is associated with HSBC Software Development (Guangdong) Limited. His patent reflects work at the intersection of machine learning, adversarial robustness, and fraud detection.

Collaborations

Bing Zhu and Ziyuan Li are identified as Xia’s coworkers. Their shared professional context connects Xia’s work with colleagues in software development.

Conclusion

Yong Xia’s patent presents a data-augmentation and adversarial-learning approach to strengthening machine-learning models for fraud detection.