Machine learning (ML) techniques play a pivotal role in modern artificial intelligence (AI) systems, but they remain notably vulnerable to disruptions caused by security attacks. These vulnerabilities can severely compromise AI system performance or be exploited maliciously, posing significant economic, ethical, and societal risks. For example, placing a small sticker on a stop sign could cause a self-driving car's perception system to misinterpret it as a speed limit sign, leading to potentially catastrophic consequences. As the reliance on AI grows, ensuring the secure, robust, and resilient operation of ML systems becomes increasingly essential. However, most robust ML research has focused on static, closed-world scenarios that fail to address the complexities of dynamic, real-world environments. This award aims to develop transformative methods to enhance the resilience and reliability of ML systems in these challenging settings. The outcome of this project promises broad societal benefits, including safer and more dependable AI applications in diverse fields such as biology, healthcare, cybersecurity, and manufacturing. Additionally, the project will transform AI education by integrating ML robustness as a foundational theme, preparing future workforce to tackle emerging challenges in trustworthy AI, and fostering public awareness of AI risks and mitigation strategies through extensive outreach. This award seeks to advance AI research by addressing three key challeng