Venus
Venus

Venus is a state-of-the-art sound and complete verification toolkit for Relu-based feed-forward neural networks. It can be used to check reachability and local adversarial robustness properties. Venus implements a MILP-based verification method whereby it leverages dependency relations between the ReLU nodes to reduce the search space that needs to be considered during branch-and-bound. It additionally implements methods based on symbolic interval propagation and input domain splitting. 

Keywords:  formal verification; local adversarial robustness; neural networks.

Acknowledgements

This work is supported in part by the  DARPA Assured Autonomy  program.

Contacts
ORGANIZATION

Imperial College London, London, United Kingdom

Contributors

Panagiotis Kouvaros

Elena Botoeva

Alessio Lomuscio

References
Botoeva, E., Kouvaros, P., Kronqvist, J., Lomuscio, A., & Misener, R. (2020). Efficient Verification of ReLU-Based Neural Networks via Dependency Analysis. Proceedings Of The Aaai Conference On Artificial Intelligence, 34, 3291-3299. http://doi.org/10.1609/aaai.v34i04.5729 (Original work published Apr.)
Kouvaros, P., & Lomuscio, A. (2018). Formal Verification of CNN-based Perception Systems. arXiv. Retrieved from https://arxiv.org/abs/1811.11373
Akintunde, M., Lomuscio, A., Maganti, L., & Pirovano, E. (2018). Reachability Analysis for Neural Agent-Environment Systems. International Conference on Principles of Knowledge Representation and Reasoning. Association for the Advancement of Artificial Intelligence. Retrieved from https://www.aaai.org/ocs/index.php/KR/KR18/paper/viewPaper/17991
Akintunde, M., Kevorchian, A., Lomuscio, A., & Pirovano, E. (2019). Verification of RNN-Based Neural Agent-Environment Systems. AAAI Conference on Artificial Intelligence. Association for the Advancement of Artificial Intelligence. http://doi.org/https://doi.org/10.1609/aaai.v33i01.33016006
Akintunde, M., Botoeva, E., Kouvaros, P., & Lomuscio, A. (2020). Formal Verification of Neural Agents in Non-deterministic Environments. International Conference on Autonomous Agents and Multiagent Systems. International Foundation for Autonomous Agents and Multiagent Systems. Retrieved from http://ifaamas.org/Proceedings/aamas2020/pdfs/p25.pdf
Henriksen, P., & Lomuscio, A. (2020). Efficient Neural Network Verification via Adaptive Refinement and Adversarial Search. European Conference on Artificial Intelligence. IOS Press. http://doi.org/https://doi.org/10.3233/FAIA200385
Akintunde, M., Botoeva, E., Kouvaros, P., & Lomuscio, A. (2020). Verifying Strategic Abilities of Neural-symbolic Multi-agent Systems. International Conference on Principles of Knowledge Representation and Reasoning. IJCAI Organization. http://doi.org/https://doi.org/10.24963/kr.2020/3
Kouvaros, P., & Lomuscio, A. (2021). Towards Scalable Complete Verification of ReLU Neural Networks via Dependency-based Branching. International Joint Conference on Artificial Intelligence. IJCAI Organization.
Batten, B., Kouvaros, P., Lomuscio, A., & Zheng, Y. (2021). Efficient Neural Network Verification via Layer-based Semidefinite Relaxations and Linear Cuts. International Joint Conference on Artificial Intelligence. IJCAI Organization.
Kouvaros, P., Kyono, T., Leofante, F., Lomuscio, A., Margineantu, D., Osipychev, D., & Zheng, Y. (2021). Formal Analysis of Neural Network-based Systems in the Aircraft Domain. In The 24th International Symposium on Formal Methods (FM21). Springer.
Lan, J., Zheng, Y., & Lomuscio, A. (2022). Tight Neural Network Verification via Semidefinite Relaxations and Linear Reformulations. The 36th AAAI Conference on Artificial Intelligence (AAAI22).