Networking, Data, and Security systems (NDSs) Group

 home

Objectives

The Networking, Data, and Security systems (NDSs) research group at McMaster University focuses on developing theories, technologies, tools and systems to address the challenges of trustworthy AI, privacy-preserving ML, smart IoT systems, emerging big data applications and their involved security and privacy concerns, therefore push the next society-transforming change.


Research Areas

  • Trustworthy AI

  • Privacy-preserving ML

  • Smart IoT Systems


    • Descriptions

    Trustworthy AI

    Deep learning has achieved remarkable performance for the analysis of visual contents. It has been reported that machine learning (ML) models can be deliberately fooled by misleading samples. With the maturity of today’s ML architecture, the shift in AI will be changed from model to trustworthyness. Whether we trust the data to train a ML model or whether we trust the result from the ML service provider is essential in developing dependable AI systems. Our recent research results to alleviate the effect of noisy labels have been published as below.


    AAAI'23 Yangdi Lu, Zhiwei Xu, and Wenbo He," Rethinking Label Refurbishment: Model Robustness under Label Noise.", in Proceedings of the the thirty-seventh AAAI Conference on Artificial Intelligence,, Washingto DC, USA, February 2023.

    NeurIPS'22 Yangdi Lu, Yang Bo and Wenbo He" Noise Attention Learning: Enhancing Noise Robustness by Gradient Scaling.", in Proceedings of the the thirty-sixth Conference on Neural Information Processing Systems (NeurIPS'22),, New Orleans, USA, December 2022.

    IJCAI'22 Yangdi Lu, Yang Bo and Wenbo He," SELC: Self-Ensemble Label Correction Improves Learning with Noisy Labels.", in Proceedings of the 31st International Joint Conference on Artifical Intelligence (IJCAI-ECAI), Vienna, Austria, July 2022.

    WSDM'22 Yangdi Lu, Yang Bo and Wenbo He," An Ensemble Model for Combating Label Noise.", in Proceedings of the theThe Fifteenth International Conference on Web Search and Data Mining (WSDM'22), Virtual, Feb. 2022.

    BigData'21 Yangdi Lu, and Wenbo He," MixNN: Combating Noisy Labels in Deep Learning by Mixing with Nearest Neighbours.", in Proceedings of the the IEEE International Conference on Big Data,, Virtual, December 2021.

    IWQoS'20 Yixin Chen, Xinye Lin, Keshi Ge, Wenbo He, and Dongsheng Li," Tag Pollution Detection in Web Videos via Cross-Modal Relevance Estimation.", in Proceedings of the IEEE/ACM International Symposium on Quality of Service (IWQoS’20), Virtual, June 2020.

    Privacy-preserving ML

    As the use of AI grows and expands and a massive amount of data are involved in training a powerful ML model, the privacy of the data providers can be implicated everywher and in unexpected ways. We believe that no signle method can address all privacy aspects alone. Our team develop privacy-preserving schemes which protect the privacy of individuals while allowing useful information and knowledge to be learned. Our preliminary work on classification over encrypted images is available:


    INFOCOM'21 Wenbo He, Shusheng Li, Wenbo Wang, Muheng Wei, and Bohua Qiu." CryptoEyes: Privacy Preserving Classification over Encrypted Images. (INFOCOM’21), Virtual, May 2021.

    Smart IoT Systems

    A Smart-IoT system learns, perceives and makes real-time decisions and predictions providing actionable insights. On the otherhand, video data is considered as biggest big data among the big data applications. Our group addresses the challenges in processing massive data from camera and/or thermal camera in an efficient and effective way. The recent publications are avaiable as follows.


    INFOCOM'21 Shusheng Li and Wenbo He,"     VideoLoc: Video-based Indoor Localization with Text Information.", in Proceedings of the the IEEE International Conference on Computer Communications (INFOCOM’21), Virtual, May 2021.

    BigData'19 Shusheng Li and Wenbo He" VidAnomaly: LSTM-autoencoder-based Adversarial Learning for One-class Video Classification with Multiple Dynamic Images.", in Proceedings of the the IEEE International Conference on Big Data, Los Angeles, CA (USA), December 2019.

    WACV'20 Yang Bo, Yangdi Lu, and Wenbo He," Few-Shot Learning of Video Action Recognition Only Based on Video Contents.", in Proceedings of the the Winter Conference on Applications of Computer Vision (WACV’20), Snowmass Village, CO (USA), March 2020.