Wenjie Wang

Title: Innovations of Wenjie Wang in Adversarial Attack Detection

Introduction

Wenjie Wang is an accomplished inventor based in Decatur, GA (US). He has made significant contributions to the field of machine learning, particularly in defending learning models against adversarial attacks. His innovative approach leverages multimodal data to enhance the robustness of learning systems.

Latest Patents

Wenjie Wang holds a patent titled "Using multimodal model consistency to detect adversarial attacks." This patent describes a method, apparatus, and computer program product designed to defend learning models that are vulnerable to adversarial example attacks. The technique recognizes that data is often available in multiple modalities, such as text and images or audio and video. By exploiting the correlations between different modalities for the same entity, this approach effectively identifies and rejects adversarial samples. If the features from one attacked modality are significantly different from those of another un-attacked modality, the system can determine that an adversarial attack is occurring. This innovative defense mechanism enhances the security of machine learning models.

Career Highlights

Wenjie Wang is currently employed at International Business Machines Corporation (IBM), where he continues to develop cutting-edge technologies in the field of artificial intelligence and machine learning. His work focuses on creating solutions that improve the reliability and security of learning models.

Collaborations

Wenjie has collaborated with notable colleagues, including Ian M Molloy and Youngja Park, who contribute to the advancement of research in their respective fields.

Conclusion

Wenjie Wang's innovative work in detecting adversarial attacks through multimodal model consistency represents a significant advancement in machine learning security. His contributions are vital for the ongoing development of robust and reliable AI systems.