An Abstraction-Refinement Approach to Verifying Convolutional Neural Networks

Convolutional neural networks have gained vast popularity due to their excellent performance in the fields of computer vision, image processing, and others. Unfortunately, it is now well known that convolutional networks often produce erroneous results — for example, minor perturbations of the inputs of these networks can result in severe classification errors. Numerous verification approaches have been proposed in recent years to prove the absence of such errors, but these are typically geared for fully connected networks and suffer from exacerbated scalability issues when applied to convolutional networks. To address this gap, we present here the Cnn-Abs framework, which is particularly aimed at the verification of convolutional networks. The core of Cnn-Abs is an abstraction-refinement technique, which simplifies the verification problem through the removal of convolutional connections in a way that soundly creates an over-approximation of the original problem; and which restores these connections if the resulting problem becomes too abstract. Cnn-Abs is designed to use existing verification engines as a backend, and our evaluation demonstrates that it can significantly boost the performance of a state-of-the-art DNN verification engine, reducing runtime by 15.7% on average.

[1]  Clark Barrett,et al.  An SMT-Based Approach for Verifying Binarized Neural Networks , 2020, TACAS.

[2]  Taylor Johnson,et al.  The Second International Verification of Neural Networks Competition (VNN-COMP 2021): Summary and Results , 2021, ArXiv.

[3]  Mykel J. Kochenderfer,et al.  Reluplex: a calculus for reasoning about deep neural networks , 2021, Formal Methods in System Design.

[4]  Zahra Rahimi Afzal,et al.  Abstraction based Output Range Analysis for Neural Networks , 2020, NeurIPS.

[5]  Kyle D. Julian,et al.  Parallelization Techniques for Verifying Neural Networks , 2020, 2020 Formal Methods in Computer Aided Design (FMCAD).

[6]  Mykel J. Kochenderfer,et al.  The Marabou Framework for Verification and Analysis of Deep Neural Networks , 2019, CAV.

[7]  Joan Bruna,et al.  Intriguing properties of neural networks , 2013, ICLR.

[8]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[9]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[10]  Chih-Hong Cheng,et al.  Maximum Resilience of Artificial Neural Networks , 2017, ATVA.

[11]  Thomas A. Henzinger,et al.  Into the unknown: Active monitoring of neural networks , 2021, RV.

[12]  Min Zhang,et al.  Tightening Robustness Verification of Convolutional Neural Networks with Fine-Grained Linear Approximation , 2021, AAAI.

[13]  Junfeng Yang,et al.  Formal Security Analysis of Neural Networks using Symbolic Intervals , 2018, USENIX Security Symposium.

[14]  Yann LeCun,et al.  The mnist database of handwritten digits , 2005 .

[15]  Cho-Jui Hsieh,et al.  Efficient Neural Network Robustness Certification with General Activation Functions , 2018, NeurIPS.

[16]  Dumitru Erhan,et al.  Going deeper with convolutions , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[17]  Russ Tedrake,et al.  Evaluating Robustness of Neural Networks with Mixed Integer Programming , 2017, ICLR.

[18]  Mykel J. Kochenderfer,et al.  Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networks , 2017, CAV.

[19]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[20]  Daniel Kroening,et al.  A survey of safety and trustworthiness of deep neural networks: Verification, testing, adversarial attack and defence, and interpretability , 2018, Comput. Sci. Rev..

[21]  Corina S. Pasareanu,et al.  DeepSafe: A Data-Driven Approach for Assessing Robustness of Neural Networks , 2018, ATVA.

[22]  Guy Katz,et al.  Towards Scalable Verification of Deep Reinforcement Learning , 2021, 2021 Formal Methods in Computer Aided Design (FMCAD).

[23]  Edmund M. Clarke,et al.  Counterexample-guided abstraction refinement , 2003, 10th International Symposium on Temporal Representation and Reasoning, 2003 and Fourth International Conference on Temporal Logic. Proceedings..

[24]  Clark W. Barrett,et al.  Provably Minimally-Distorted Adversarial Examples , 2017 .

[25]  Michael Schapira,et al.  Verifying Deep-RL-Driven Systems , 2019, NetAI@SIGCOMM.

[26]  Mykel J. Kochenderfer,et al.  Global Optimization of Objective Functions Represented by ReLU Networks , 2021, Machine Learning.

[27]  Aditya V. Thakur,et al.  Correcting Deep Neural Networks with Small, Generalizing Patches , 2019 .

[28]  Jun Zhao,et al.  Recurrent Convolutional Neural Networks for Text Classification , 2015, AAAI.

[29]  Jan Kretínský,et al.  DeepAbstract: Neural Network Abstraction for Accelerating Verification , 2020, ATVA.

[30]  Thomas A. Henzinger,et al.  Formal Methods with a Touch of Magic , 2020, 2020 Formal Methods in Computer Aided Design (FMCAD).

[31]  Ori Lahav,et al.  Pruning and Slicing Neural Networks using Formal Verification , 2021, 2021 Formal Methods in Computer Aided Design (FMCAD).

[32]  Min Wu,et al.  Safety Verification of Deep Neural Networks , 2016, CAV.

[33]  Jürgen Schmidhuber,et al.  Multi-column deep neural networks for image classification , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[34]  Dahua Lin,et al.  Fastened CROWN: Tightened Neural Network Robustness Certificates , 2019, AAAI.

[35]  Alessio Lomuscio,et al.  An approach to reachability analysis for feed-forward ReLU neural networks , 2017, ArXiv.

[36]  Caterina Urban,et al.  Perfectly parallel fairness certification of neural networks , 2020, Proc. ACM Program. Lang..

[37]  Yannic Noller,et al.  NNrepair: Constraint-based Repair of Neural Network Classifiers , 2021, CAV.

[38]  Ashish Tiwari,et al.  Output Range Analysis for Deep Neural Networks , 2017, ArXiv.

[39]  Isil Dillig,et al.  Optimization and abstraction: a synergistic approach for analyzing neural network robustness , 2019, PLDI.

[40]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[41]  Antonio Criminisi,et al.  Measuring Neural Net Robustness with Constraints , 2016, NIPS.

[42]  Sijia Liu,et al.  CNN-Cert: An Efficient Framework for Certifying Robustness of Convolutional Neural Networks , 2018, AAAI.

[43]  Swarat Chaudhuri,et al.  AI2: Safety and Robustness Certification of Neural Networks with Abstract Interpretation , 2018, 2018 IEEE Symposium on Security and Privacy (SP).

[44]  Cho-Jui Hsieh,et al.  A Convex Relaxation Barrier to Tight Robustness Verification of Neural Networks , 2019, NeurIPS.

[45]  Alessio Lomuscio,et al.  Formal Verification of CNN-based Perception Systems , 2018, ArXiv.

[46]  Shweta Shinde,et al.  Quantitative Verification of Neural Networks and Its Security Applications , 2019, CCS.

[47]  Inderjit S. Dhillon,et al.  Towards Fast Computation of Certified Robustness for ReLU Networks , 2018, ICML.

[48]  John Schulman,et al.  Concrete Problems in AI Safety , 2016, ArXiv.

[49]  Radu Calinescu,et al.  DeepCert: Verification of Contextually Relevant Robustness for Neural Network Image Classifiers , 2021, SAFECOMP.

[50]  Nathan Srebro,et al.  Equality of Opportunity in Supervised Learning , 2016, NIPS.

[51]  Timon Gehr,et al.  An abstract domain for certifying neural networks , 2019, Proc. ACM Program. Lang..

[52]  Weiming Xiang,et al.  Verification of Deep Convolutional Neural Networks Using ImageStars , 2020, CAV.

[53]  P. Cochat,et al.  Et al , 2008, Archives de pediatrie : organe officiel de la Societe francaise de pediatrie.

[54]  Guy Katz,et al.  Minimal Modifications of Deep Neural Networks using Verification , 2020, LPAR.

[55]  R. Sarpong,et al.  Bio-inspired synthesis of xishacorenes A, B, and C, and a new congener from fuscol† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c9sc02572c , 2019, Chemical science.

[56]  Guy Katz,et al.  Minimal Multi-Layer Modifications of Deep Neural Networks , 2021, NSV/FoMLAS@CAV.

[57]  Justin Emile Gottschlich,et al.  An Abstraction-Based Framework for Neural Network Verification , 2019, CAV.

[58]  Christian Tjandraatmadja,et al.  Strong mixed-integer programming formulations for trained neural networks , 2018, Mathematical Programming.

[59]  Taylor T. Johnson,et al.  Reachability Analysis of Convolutional Neural Networks , 2021, ArXiv.

[60]  Corina S. Pasareanu,et al.  Automated Assume-Guarantee Reasoning by Abstraction Refinement , 2008, CAV.

[61]  Rüdiger Ehlers,et al.  Formal Verification of Piece-Wise Linear Feed-Forward Neural Networks , 2017, ATVA.

[62]  Krishnendu Chatterjee,et al.  Run-Time Optimization for Learned Controllers Through Quantitative Games , 2019, CAV.

[63]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[64]  Jin Xu.  Conv-Reluplex : A Verification Framework For Convolution Neural Networks (S) , 2021, Proceedings of the 33rd International Conference on Software Engineering and Knowledge Engineering.

[65]  Clark W. Barrett,et al.  Simplifying Neural Networks Using Formal Verification , 2020, NFM.

[66]  Nham Le,et al.  Verification of Recurrent Neural Networks for Cognitive Tasks via Reachability Analysis , 2020, ECAI.

[67]  Guy Katz,et al.  Verifying Recurrent Neural Networks using Invariant Inference , 2020, ATVA.

[68]  Guy Katz,et al.  Verifying learning-augmented systems , 2021, SIGCOMM.

[69]  Mykel J. Kochenderfer,et al.  Toward Scalable Verification for Safety-Critical Deep Networks , 2018, ArXiv.

[70]  Mykel J. Kochenderfer,et al.  Towards Proving the Adversarial Robustness of Deep Neural Networks , 2017, FVAV@iFM.

[71]  Ekaterina Komendantskaya,et al.  Property-driven Training: All You (N)Ever Wanted to Know About , 2021, ArXiv.