Timezone: »

 
Certified Defences Against Adversarial Patch Attacks on Semantic Segmentation
Maksym Yatsura · Kaspar Sakmann · N. Grace Hua · Matthias Hein · Jan Hendrik Metzen

@
in Trustworthy and Socially Responsible Machine Learning »
Event URL: https://openreview.net/forum?id=VFGgG8XpFLu »

Adversarial patch attacks are an emerging security threat for real world deep learning applications. We present Demasked Smoothing, the first approach (up to our knowledge) to certify the robustness of semantic segmentation models against this threat model. Previous work on certifiably defending against patch attacks has mostly focused on image classification task and often required changes in the model architecture and additional training which is undesirable and computationally expensive. In Demasked Smoothing, any segmentation model can be applied without particular training, fine-tuning, or restriction of the architecture. Using different masking strategies, Demasked Smoothing can be applied both for certified detection and certified recovery. In extensive experiments we show that Demasked Smoothing can on average certify 63% of the pixel predictions for a 1% patch in the detection task and 46% against a 0.5% patch for the recovery task on the ADE20K dataset.

[ Poster]  [ OpenReview]  [ Topia] 

Author Information

Maksym Yatsura (Bosch Center for Artificial Intelligence)
Kaspar Sakmann (Bosch Center for Artificial Intelligence)
N. Grace Hua (Robert Bosch GmbH, Bosch)
Matthias Hein (University of Tübingen)
Jan Hendrik Metzen (Robert Bosch GmbH)

More from the Same Authors