Timezone: »

Certifying Geometric Robustness of Neural Networks
Mislav Balunovic · Maximilian Baader · Gagandeep Singh · Timon Gehr · Martin Vechev

Thu Dec 12 10:45 AM -- 12:45 PM (PST) @ East Exhibition Hall B + C #13

The use of neural networks in safety-critical computer vision systems calls for their robustness certification against natural geometric transformations (e.g., rotation, scaling). However, current certification methods target mostly norm-based pixel perturbations and cannot certify robustness against geometric transformations. In this work, we propose a new method to compute sound and asymptotically optimal linear relaxations for any composition of transformations. Our method is based on a novel combination of sampling and optimization. We implemented the method in a system called DeepG and demonstrated that it certifies significantly more complex geometric transformations than existing methods on both defended and undefended networks while scaling to large architectures.

Author Information

Mislav Balunovic (ETH Zurich)
Maximilian Baader (ETH Zürich)
Gagandeep Singh (ETH Zurich)
Timon Gehr (ETH Zurich)
Martin Vechev (ETH Zurich, Switzerland)

More from the Same Authors