for security and
play

for Security and Services Antonio Rizzo, Alessandro Rossi, - PowerPoint PPT Presentation

CyberPhysical systems for Security and Services Antonio Rizzo, Alessandro Rossi, Francesco Montefoschi, Giovanni Burresi Carlo Festucci, Maurizio Caporali Siena, 14 Sett 2018 Case of study: ATM-Sense TREND Rapine vs Attacchi ATM - Fonte


  1. CyberPhysical systems for Security and Services Antonio Rizzo, Alessandro Rossi, Francesco Montefoschi, Giovanni Burresi Carlo Festucci, Maurizio Caporali Siena, 14 Sett 2018

  2. Case of study: ATM-Sense

  3. TREND Rapine vs Attacchi ATM - Fonte ABI 2017 Numero Rapine per Anno 1.400 1.200 1.000 800 600 400 200 0 2012 2013 2014 2015 2016

  4. ATM attacks

  5. New Attacks https://www.europol.europa.eu/newsroom/news/27-arrested-in-successful-hit-against-atm-black-box-attacks

  6. Video Surveillance Approach

  7. Intel RealSense Depth Cameras • Powerful Open Souce SDK • Easily Embeddable

  8. Intel RealSense Depth Cameras

  9. Convolutional Neural Networks maxpool maxpool convolutions source image convolutions fully connected

  10. Image Convolutions

  11. Image Convolutions

  12. Image Convolutions

  13. Convolutions

  14. Max Pooling maxpool source image convolutions

  15. More layers... maxpool maxpool convolutions source image convolutions

  16. Visualizing Convolutional Layers References: - Lee, H., Grosse, R., Ranganath, R., & Ng, A. Y. (2009, June). Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations. In Proceedings of the 26th annual international conference on machine learning (pp. 609-616). ACM.

  17. Convolutional Neural Networks maxpool maxpool convolutions source image convolutions fully connected

  18. CNN: ImageNet Classification Error References: - Russakovsky, Olga, et al. "Imagenet large scale visual recognition challenge." International Journal of Computer Vision 115.3 (2015): 211-252 - Hardware Architectures for Deep Neural Networks, ISCA Tutorial, MIT

  19. Machine Learning Process 1. Get a dataset 2. Define the network architecture 3. Train and Test the model

  20. 1. Get a Dataset

  21. 2. Define the Network Architecture

  22. 3. Train and Test the Model

  23. Results Single Frame analysis: Five Frames analysis: ● No false alarms Test Dataset Classification Accuracy ● No undetected attacks ● Attack detection time: Background 98.19% ○ mean: 2.4 sec Withdrawal 97.05% ○ max: 3.3 sec Attack 98.32% Average 97.85% ○ mean: 0.5 sec Model Running on SECO SBC – A80 with Intel Braswell CPU

  24. Predicting Security Thank You

Recommend


More recommend