RoBIC: A benchmark suite for assessing classifiers robustness

Thibault Maho (1) , Benoît Bonnet (1) , Teddy Furon (1) , Erwan Le Merrer (2)

1 LinkMedia - Creating and exploiting explicit links between multimedia fragments
2 WIDE - the World Is Distributed Exploring the tension between scale and coordination

Many defenses have emerged with the development of adversarial attacks. Models must be objectively evaluated accordingly. This paper systematically tackles this concern by proposing a new parameter-free benchmark we coin RoBIC. RoBIC fairly evaluates the robustness of image classifiers using a new half-distortion measure. It gauges the robustness of the network against white and black box attacks, independently of its accuracy. RoBIC is faster than the other available benchmarks. We present the significant differences in the robustness of 16 recent models as assessed by RoBIC. We make this benchmark publicly available for use and contribution at https://gitlab.inria.fr/tmaho/ robustness_benchmark.

Mots clés: Benchmark; adversarial examples; model robustness; half-distortion measure

Domaines: Machine Learning [stat.ML]

Thibault Maho, Benoît Bonnet, Teddy Furon, Erwan Le Merrer. RoBIC: A benchmark suite for assessing classifiers robustness. ICIP 2021 - IEEE International Conference on Image Processing, Sep 2021, Anchorage, Alaska, United States. pp.1-5, ⟨10.1109/ICIP42928.2021.9506053⟩. ⟨hal-03234791⟩