Rise from the Ashes: LLM-Based Static Analysis for Deep Learning Framework Bugs

Deep learning DL frameworks are critical AI infrastructures that often hide bugs with serious security implications. While dynamic approaches such as fuzzing are effective in uncovering these bugs, they require real test execution and incur high computational costs. Static analysis is a natural...