This course aims to teach learners about explaining deep learning-based security applications. The learning outcomes include understanding the concerns of opaque deep learning models, exploring existing explanation techniques and their limitations, learning about the LEMNA method for local explanations, and building trust in target models. The course covers skills such as modeling feature dependency, deriving explanations from deep neural networks using LEMNA, and evaluating explanation accuracy. The teaching method includes theoretical explanations, demonstrations, and troubleshooting sessions. The intended audience for this course includes individuals interested in deep learning, security applications, and model interpretability.
LEMNA - Explaining Deep Learning Based Security Applications
Association for Computing Machinery (ACM) via YouTube
Overview
Syllabus
Intro
The Concerns of Opaque Deep Learning Model
Existing Explanation Techniques & Limitations
One Example of Model Explanation (LIME. KDD'16)
Limitation of Existing Explanation Techniques
LEMNA: Local Explanation Method using Nonlinear Approximation
Supporting Locally Non-linear Decision Boundaries
Modeling the Feature Dependency . Mature regression model with fused lasso
Deriving an Explanation from DNN with LEMNA
Explanation Accuracy Evaluation
Demonstration of LEMNA in Identifying Binary Function Start
Building Trust in the Target Models
Troubleshooting and Patching Model Errors
Taught by
Association for Computing Machinery (ACM)
