Compositional Falsification of Cyber-Physical Systems with Machine Learning Components


Cyber-physical systems (CPS), such as automotive systems, are starting to include sophisticated machine learning (ML) components. Their correctness, therefore, depends on properties of the inner ML modules. While learning algorithms aim to generalize from examples, they are only as good as the examples provided, and recent efforts have shown that they can produce inconsistent output under small adversarial perturbations. This raises the question: can the output from learning components lead to a failure of the entire CPS? In this work, we address this question by formulating it as a problem of falsifying signal temporal logic specifications for CPS with ML components. We propose a compositional falsification framework where a temporal logic falsifier and a machine learning analyzer cooperate with the aim of finding falsifying executions of the considered model. The efficacy of the proposed technique is shown on an automatic emergency braking system model with a perception component based on deep neural networks.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11


  1. 1.

  2. 2.

  3. 3.

    Udacity’s Self-Driving Car Simulator:

  4. 4.


  5. 5.

    Socket.IO protocol:


  1. 1.


  2. 2.

    Udacity self-driving car simulator built with unity.

  3. 3.

    Abadi, M. et al.: TensorFlow: Large-scale machine learning on heterogeneous systems (2015). Software available from

  4. 4.

    Annpureddy, Y., Liu, C., Fainekos, G.E., Sankaranarayanan, S.: S-taliro: a tool for temporal logic falsification for hybrid systems. In: Tools and Algorithms for the Construction and Analysis of Systems, TACAS, pp. 254–257 (2011)

    Google Scholar 

  5. 5.

    Blum, A.L., Langley, P.: Selection of relevant features and examples in machine learning. Artif. Intell. 97(1), 245–271 (1997)

    MathSciNet  Article  Google Scholar 

  6. 6.

    Bojarski, M., Del Testa, D., Dworakowski, D., Firner, B., Flepp, B., Goyal, P., Jackel, L.D., Monfort, M., Muller, U., Zhang, J., et al.: End to end learning for self-driving cars (2016). arXiv preprint arXiv:1604.07316

  7. 7.

    Branicky, M.S., LaValle, S.M., Olson, K., Yang, L.: Quasi-randomized path planning. In: IEEE International Conference on Robotics and Automation, 2001. Proceedings 2001 ICRA, vol. 2, pp. 1481–1487. IEEE (2001)

  8. 8.

    Carlini, N., Wagner, D.: Towards evaluating the robustness of neural networks. In: 2017 IEEE Symposium on Security and Privacy (SP), pp. 39–57 (2017)

  9. 9.

    Donzé, A.: Breach, a toolbox for verification and parameter synthesis of hybrid systems. In: Computer Aided Verification, CAV, pp. 167–170 (2010)

    Google Scholar 

  10. 10.

    Donzé, A., Ferrere, T., Maler, O.: Efficient robust monitoring for STL. In: Computer Aided Verification, CAV, pp. 264–279. Springer, Berlin (2013)

    Google Scholar 

  11. 11.

    Dreossi, T., Dang, T., Donzé, A., Kapinski, J., Jin, X., Deshmukh, J.: Efficient guiding strategies for testing of temporal properties of hybrid systems. In: NASA Formal Methods, NFM, pp. 127–142 (2015)

    Google Scholar 

  12. 12.

    Dreossi, T., Donzé, A., Seshia, S.A.: Compositional falsification of cyber-physical systems with machine learning components. In: NASA Formal Methods Conference (NFM) (2017)

  13. 13.

    Dreossi, T., Ghosh, S., Sangiovanni-Vincentelli, A.L., Seshia, S.A.: Systematic testing of convolutional neural networks for autonomous driving. In: ICML Workshop on Reliable Machine Learning in the Wild (RMLW) (2017). arXiv:1708.03309

  14. 14.

    Dreossi, T., Jha, S., Seshia, S.A.: Semantic adversarial deep learning. In: 30th International Conference on Computer Aided Verification (CAV) (2018)

    Google Scholar 

  15. 15.

    Duggirala, P.S., Mitra, S., Viswanathan, M., Potok, M.: C2E2: a verification tool for stateflow models. In: International Conference on Tools and Algorithms for the Construction and Analysis of Systems, pp. 68–82. Springer, Berlin (2015)

    Google Scholar 

  16. 16.

    Fawzi, A., Fawzi, O., Frossard, P.: Analysis of classifiers’ robustness to adversarial perturbations (2015). arXiv preprint arXiv:1502.02590

  17. 17.

    Hannaford, B.: Resolution-first scanning of multidimensional spaces. CVGIP Graph. Models Image Process. 55(5), 359–369 (1993)

    Article  Google Scholar 

  18. 18.

    Hinton, G., et al.: Deep neural networks for acoustic modeling in speech recognition: the shared views of four research groups. IEEE Signal Process. Mag. 29(6), 82–97 (2012)

    Article  Google Scholar 

  19. 19.

    Huang, X., Kwiatkowska, M., Wang, S., Wu, M.: Safety verification of deep neural networks (2016). CoRR arXiv:1610.06940

  20. 20.

    Iandola, F.N., Han, S., Moskewicz, M.W., Ashraf, K., Dally, W.J., Keutzer, K.: Squeezenet: Alexnet-level accuracy with 50x fewer parameters and \(<\) 0.5 mb model size (2016). arXiv preprint arXiv:1602.07360

  21. 21.

    Jia, Y., Shelhamer, E., Donahue, J., Karayev, S., Long, J., Girshick, R., Guadarrama, S., Darrell, T.: Caffe: convolutional architecture for fast feature embedding. In: ACM Multimedia Conference, ACMMM, pp. 675–678 (2014)

  22. 22.

    Jin, X., Donzé, A., Deshmukh, J., Seshia, S.A.: Mining requirements from closed-loop control models. IEEE Trans. Comput.-Aided Des. Circuits Syst. 34(11), 1704–1717 (2015)

    Article  Google Scholar 

  23. 23.

    Krizhevsky, A., Sutskever, I., Hinton, G.E.: Imagenet classification with deep convolutional neural networks. In: Advances in Neural Information Processing Systems, pp. 1097–1105 (2012)

  24. 24.

    Maler, O., Nickovic, D.: Monitoring temporal properties of continuous signals. In: Formal Techniques, Modelling and Analysis of Timed and Fault-Tolerant Systems, pp. 152–166. Springer, Berlin (2004)

    Google Scholar 

  25. 25.

    Matousek, J.: Geometric Discrepancy: An Illustrated Guide, vol. 18. Springer, Berlin (2009)

    Google Scholar 

  26. 26.

    Michalski, R .S., Carbonell, J .G., Mitchell, T .M.: Machine Learning: An Artificial Intelligence Approach. Springer, Berlin (2013)

    Google Scholar 

  27. 27.

    Moosavi-Dezfooli, S.-M., Fawzi, A., Frossard, P.: DeepFool: a simple and accurate method to fool deep neural networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2574–2582 (2016)

  28. 28.

    Morokoff, W.J., Caflisch, R.E.: Quasi-random sequences and their discrepancies. SIAM J. Sci. Comput. 15(6), 1251–1279 (1994)

    MathSciNet  Article  Google Scholar 

  29. 29.

    Nguyen, A., Yosinski, J., Clune, J.: Deep neural networks are easily fooled: high confidence predictions for unrecognizable images. In: Computer Vision and Pattern Recognition, CVPR, pp. 427–436. IEEE (2015)

  30. 30.

    Niederreiter, H.: Low-discrepancy and low-dispersion sequences. J. Number Theory 30(1), 51–70 (1988)

    MathSciNet  Article  Google Scholar 

  31. 31.

    Niederreiter, H.: Random Number Generation and Quasi-Monte Carlo Methods. SIAM, Philadelphia (1992)

    Google Scholar 

  32. 32.

    Pei, K., Cao, Y., Yang, J., Jana, S.: DeepXplore: automated whitebox testing of deep learning systems. In: Proceedings of the 26th Symposium on Operating Systems Principles (SOSP), pp. 1–18 (2017)

  33. 33.

    Rosenblatt, J., Wierdl, M.: Pointwise ergodic theorems via harmonic analysis. In: Conference on Ergodic Theory, No. 205, pp. 3–151 (1995)

  34. 34.

    Seshia, S.A., Desai, A., Dreossi, T., Fremont, D.J., Ghosh, S., Kim, E., Shivakumar, S., Vazquez-Chanlatte, M., Yue, X.: Formal specification for deep neural networks. In: 16th International Symposium on Automated Technology for Verification and Analysis (ATVA), pp. 20–34 (2018)

    Google Scholar 

  35. 35.

    Seshia, S.A., Sadigh, D., Sastry, S.S.: Towards verified artificial intelligence (2016). CoRR arXiv:1606.08514

  36. 36.

    Shirley, P. et al.: Discrepancy as a quality measure for sample distributions. In: Proceedings of Eurographics, vol. 91, pp. 183–194 (1991)

  37. 37.

    Sloan, I .H., Joe, S.: Lattice Methods for Multiple Integration. Oxford University Press, Oxford (1994)

    Google Scholar 

  38. 38.

    Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I., Fergus, R.: Intriguing properties of neural networks (2013). arXiv:1312.6199

  39. 39.

    Taeyoung, L., Kyongsu, Y., Jangseop, K., Jaewan, L.: Development and evaluations of advanced emergency braking system algorithm for the commercial vehicle. In: Enhanced Safety of Vehicles Conference, ESV, pp. 11–0290 (2011)

  40. 40.

    Trandafir, Aurel., Weisstein, Eric, W.: Quasirandom sequence. From MathWorld—A Wolfram Web Resource

  41. 41.

    Vapnik, V.: Principles of risk minimization for learning theory. In: NIPS, pp. 831–838 (1991)

  42. 42.

    Vazquez-Chanlatte, M., Deshmukh, J.V., Jin, X., Seshia, S.A.: Logical clustering and learning for time-series data. In: Computer Aided Verification—29th International Conference (CAV), pp. 305–325 (2017)

    Google Scholar 

  43. 43.

    Weyl, H.: Über die gleichverteilung von zahlen mod. eins. Math. Ann. 77(3), 313–352 (1916)

    MathSciNet  Article  Google Scholar 

  44. 44.

    Yamaguchi, T., Kaga, T., Donzé, A., Seshia, S.A.: Combining requirement mining, software model checking, and simulation-based verification for industrial automotive systems. In: Proceedings of the IEEE International Conference on Formal Methods in Computer-Aided Design (FMCAD) (2016)

Download references

Author information



Corresponding author

Correspondence to Tommaso Dreossi.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This work is funded in part by the DARPA BRASS program under Agreement Number FA8750-16-C-0043, NSF Grants CNS-1646208, CNS-1545126, CCF-1837132, and CCF-1139138, the DARPA Assured Autonomy program, Toyota under the iCyPhy center, Berkeley Deep Drive, and by TerraSwarm, one of six centers of STARnet, a Semiconductor Research Corporation program sponsored by MARCO and DARPA. All the contributions of the second author with the exception of those to Sect. 5.2 occurred while he was affiliated with UC Berkeley. We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Dreossi, T., Donzé, A. & Seshia, S.A. Compositional Falsification of Cyber-Physical Systems with Machine Learning Components. J Autom Reasoning 63, 1031–1053 (2019).

Download citation


  • Cyber-physical systems
  • Machine learning
  • Falsification
  • Temporal logic
  • Deep learning
  • Neural networks
  • Autonomous driving