Training Description
Key Takeaways
- Focuses on secure, accountable, and verifiable machine learning
- Explores privacy, robustness, and fairness in ML systems
- Addresses software testing and validation for ML, including federated learning
- Brings together academic and industry experts
- Targets researchers, engineers, and security professionals
SAFE-ML 2026 is the 2nd International Workshop on Secure, Accountable, and Verifiable Machine Learning, co-located with the 19th IEEE International Conference on Software Testing, Verification and Validation. This event is dedicated to advancing the security, reliability, and trustworthiness of machine learning systems as they become increasingly vital across industries. The workshop provides a platform for experts to discuss innovative solutions for the challenges facing modern ML deployments.
Workshop Focus and Themes
The workshop centers on the intersection of machine learning and software testing. Key topics include privacy preservation, adversarial robustness, and mitigation of bias in ML models. Attendees will explore security against poisoning attacks, unlearning algorithms, and privacy for large language models. The event also highlights secure aggregation and updates in federated learning, as well as robustness against malicious clients.
Technical approaches such as fuzzing, out-of-distribution detection, and membership inference attacks are discussed. The workshop encourages the development of methods and tools that ensure ML systems are safe, reliable, and accountable for real-world deployment.
Target Audience and Relevance
SAFE-ML 2026 is designed for researchers, practitioners, and professionals in software testing, machine learning, cybersecurity, and data privacy. The audience includes software engineers, ML engineers, security analysts, academic researchers, and technical leads from organizations deploying or testing ML systems. The workshop is especially relevant for those working in safety-critical or regulated industries where trust and compliance are paramount.
Addressing Critical Challenges
The event addresses pressing business and technical challenges such as ensuring the security and privacy of ML models, detecting and mitigating bias, and validating robustness against adversarial attacks. It also covers verifying data integrity in training sets and providing accountability and explainability in ML-driven decision-making. By fostering collaboration between academia and industry, the workshop aims to advance the deployment of trustworthy and resilient ML systems.
Format and Experience
SAFE-ML 2026 is held in person in Daejeon, South Korea. The workshop features research paper presentations, expert-led discussions, and networking opportunities. The event is technical and research-focused, providing a forum for thought leadership and community building in the field of secure and trustworthy machine learning.