CRJul 18, 2016
On the Effectiveness of Defensive Distillation
arXiv:1607.05113v167 citations
Originality Synthesis-oriented
AI Analysis
This addresses security vulnerabilities in machine learning models for practitioners, though it is incremental as it extends an existing defense to more attacks.
The study found that defensive distillation effectively reduces adversarial samples from both the fast gradient sign method and the Jacobian-based iterative attack, with experimental results showing mitigation success.
We report experimental results indicating that defensive distillation successfully mitigates adversarial samples crafted using the fast gradient sign method, in addition to those crafted using the Jacobian-based iterative attack on which the defense mechanism was originally evaluated.