Event schedules / ICONIP 2026

SS-04: Reliable, Robust and Secure Machine Learning Algorithms

Monday, November 23, 2026, – (Australia/Sydney)
Clarendon Room E · Special Session

Reliable, Robust and Secure Machine Learning Algorithms

This session addresses the critical need for designing and deploying trustworthy AI systems in high-stakes domains like healthcare, finance, and autonomous driving. It focuses on enhancing the robustness, performance guarantees, and safety of machine learning algorithms against threats such as adversarial attacks, data shifts, and model misspecifications. Key topics include bias and fairness, explainability, privacy-preserving model building, and the assessment of Large Language Models (LLMs). Building on a successful history at ICONIP since 2021, the session serves as a forum for academic and industrial experts to exchange theoretical and empirical advances in creating secure and reliable intelligent systems.Invited Speaker | Professor Yong XiangDeep learning-based image watermarking against forgery attacks

Paper 108 - Raw Top-2 Margin Calibration for Multiclass Selective Classification on Public Tabular Benchmarks
Authors: Zhipeng Chen, Haolun Tang, Jingyi Zhan and Yan Feng

Paper 172 - Mitigating Toxicity Drift in Speech Recognition via Multi-Level Semantic Alignment
Authors: Dan Kienast and Jiaojiao Jiang

Paper 265 - HybridSplit: Efficient Homomorphic Encryption-Based Split Inference via Native Polynomial Networks
Authors: Dongfeng Huang, Zhicheng Xu, Hongbing Cheng, Huijun Zhang and Lu Zhou