CLEVER score

{{Orphan|date=September 2018}}

{{notability|date=September 2018}}

The CLEVER (Cross Lipschitz Extreme Value for nEtwork Robustness) score is a way of measuring the robustness of an artificial neural network towards adversarial attacks.{{cite arXiv |last=Weng |first=Tsui-Wei |date=2018 |title=Evaluating the robustness of neural networks: An extreme value theory approach |class=stat.ML |eprint=1801.10578}}

It was developed by a team at the MIT-IBM Watson AI Lab in IBM Research and first presented at the 2018 International Conference on Learning Representations.{{cite web |url=https://www.ibm.com/blogs/research/2018/05/clever-adversarial-attack/ |title=A CLEVER Way to Resist Adversarial Attack |last= |first= |date=May 2, 2018 |website= IBM|publisher= |access-date=September 12, 2018 |quote=}} It was mentioned and reviewed by Ian Goodfellow{{cite web|url=https://openreview.net/forum?id=BkUHlMZ0b |title=Evaluating the Robustness of Neural Networks: An Extreme Value Theory Approach|date=10 February 2022 }} as well. It was adopted into an educational game Fool The Bank{{ cite web|url=https://foolthebank.mybluemix.net/ |title=Fool the Bank - IBM Research}} by Narendra Nath Joshi,{{cite web|url=http://nnjoshi.co |title=Narendra Nath Joshi}} Abhishek Bhandwaldar and Casey Dugan