Taller Cybercamp 2017. Seguridad ofensiva y defensiva con Machine Learning

Dr. Alfonso Muñoz - Head of Cybersecurity lab (@mindcrypt)

D. José Ignacio Escribano - Security & Machine Learning Researcher (joseignacio.escribano@i4s.com)

Cybersecurity Lab - Innovation 4 Security (BBVA group)

Seguridad defensiva

Papers

• Learning from Simulated and Unsupervised Images through Adversarial Training
• Steganographic Generative Adversarial Networks
• Learning to Protect Communications with Adversarial Neural Cryptography
• SSGAN: Secure Steganography Based on Generative Adversarial Networks
• Generating steganographic images via adversarial training
• CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy
• Privacy-Preserving Classification on Deep Neural Network

Proyectos

• MLSec Project

Seguridad ofensiva

Herramientas/Librerías

• Cleverhans
• FloydHub
• AdversariaLib
• AlfaSVMLib

Adversarial Machine Learning

• Deep Learning Adversarial Examples – Clarifying Misconceptions
• Awesome Adversarial Machine Learning
• Introduction to Adversarial Machine Learning

Ataques

• Explaining and Harnessing Adversarial Examples
• The Limitations of Deep Learning in Adversarial Settings
• Practical Black-Box Attacks against Machine Learning
• BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
• PassGAN: A Deep Learning Approach for Password Guessing
• Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition
• Stealing Machine Learning Models via Prediction APIs
• Data Driven Exploratory Attacks on Black Box Classifiers in Adversarial Domains
• Distillation as a Defense to Adversarial Perturbations against Deep Neural Networks
• Universal adversarial perturbations
• Adversarial examples in the physical world
• Poisoning attacks against SVMs

Demos

• A Neural Network Playground - TensorFlow
• TensorFire
• Transferencia de estilo
• Generación de contraseñas con Keras
• Criptografía
• Esteganografía
• Cleverhans
• Deep Pwning

Recursos

Código

• Adversarial Steganography
• Neural Cryptography Tensorflow
• Adversarial Neural Crypt
• Deep Pwning
• Cleverhans
• Generación de contraseñas
• Transferencia de estilo

Workshops

• Machine Learning and Computer Security Workshop
• ACM Workshop on Artificial Intelligence and Security

Repositorios

Repositorios de datos

• KAPSARC Data Portal
• Datasets | Kaggle
• US Government Web Services and XML Data Sources
• AWS Public Datasets
• Registry of Research Data Repositories
• Reddit datasets
• Samples of Security Related Data
• SecLists
• List of datasets for machine learning research

Bases de datos de imágenes

• THE MNIST DATABASE of handwritten digits
• CIFAR-10 and CIFAR-100 datasets
• Large-scale CelebFaces Attributes (CelebA) Dataset
• ImageNet
• COCO - Common Objects in Context

Librerías/Frameworks de Machine Learning/Deep Learning

• scikit-learn: Machine Learning in Python
• H2O.ai
• MLlib | Apache Spark
• Weka
• Apache Mahout: Scalable machine learning and data mining
• TensorFlow
• Caffe 2 | A New Lightweight, Modular, and Scalable Deep Learning Framework
• Torch
• Theano
• Keras

Servicios en la nube

• Inteligencia artificial en AWS. Aprendizaje automático potente para desarrolladores y científicos de datos
• TensorFlow en AWS – Aprendizaje profundo en la nube
• Amazon Machine Learning - Predictive Analytics with AWS
• Google Cloud Machine Learning at Scale | Google Cloud Platform
• Cloud Machine Learning Engine - Google Cloud Platform
• Machine Learning | Microsoft Azure
• FloydHub - Deep Learning Platform - Cloud GPU
• BigML.com is Machine Learning made easy

Miscelánea

• Moral Machine
• RoboLaw
• Artificial Intelligence Safety and Cybersecurity: a Timeline of AI Failures
• An Open Letter RESEARCH PRIORITIES FOR ROBUST AND BENEFICIAL ARTIFICIAL INTELLIGENCE
• Microsoft deletes 'teen girl' AI after it became a Hitler-loving sex robot within 24 hours
• “Racist” Camera Phenomenon Explained — Almost
• Google Apologizes For Tagging Photos Of Black People As ‘Gorillas’
• When Will AI Exceed Human Performance? Evidence from AI Experts
• Killer robots? Musk and Zuckerberg escalate row over dangers of AI
• Researchers warn against the rise of "big data hubris"
• Ciencia de datos o augurio: El método científico en la era del Big Data
• Artículo 22 UE RGDP "Decisiones individuales automatizadas, incluida la elaboración de perfiles"
• THE NEURAL NETWORK ZOO