Unleashing the Power of Decompilation for x86 Deep Neural Network Executables

Unleashing the Power of Decompilation for x86 Deep Neural Network Executables

Explore a groundbreaking 24-minute Black Hat conference talk on BTD (Bin to DNN), a decompiler for deep neural network (DNN) executables. Delve into the world of deep learning models compiled into executables for heterogeneous hardware devices. Discover how BTD leverages low-level hardware primitives to enable efficient DL computations across various computing platforms, including CPUs, GPUs, and hardware accelerators. Learn about the innovative approach presented by researchers Tianxiang Li, Wenqiang Li, Zhibo Liu, Shuai Wang, Xiaofei Xie, and Yuanyuan Yuan to decompile DNN executables, potentially revolutionizing the field of deep learning and cybersecurity.