Presented By O’Reilly and Intel Nervana
Put AI to work
September 17-18, 2017: Training
September 18-20, 2017: Tutorials & Conference
San Francisco, CA

Deep learning in enterprise IoT: Use cases and challenges

Jisheng Wang (Aruba, a Hewlett Packard Enterprise Company)
4:50pm–5:30pm Tuesday, September 19, 2017
Verticals and applications
Location: Imperial A Level: Intermediate
Secondary topics:  Deep learning, IoT (including smart cities, manufacturing, smart homes/buildings)
Average rating: *****
(5.00, 1 rating)

What you'll learn

  • Explore two real-world enterprise IoT problems—device identification and IoT security—and examples of applying deep learning to solving some real-world non-image or non-NLP problems
  • Understand the challenges of and solutions for applying deep learning in enterprise applications

Description

Recently, both deep learning and the IoT have attracted tremendous attention. Jisheng Wang shares firsthand experience in applying deep learning to solving some real-world enterprise IoT problems (e.g., IoT device identification and IoT security) and outlines some challenges for deep learning in enterprise applications, along with suggestions to overcome them.

Jisheng covers two enterprise IoT use case studies: IoT device identification and behavior anomaly detection for enterprise IoT devices. Given the variety and volume of emerging IoT devices in the near future, identifying and managing IoT devices automatically is the biggest challenge to enterprise IT. Jisheng presents a recurrent neural network (RNN)-based approach to automatically identify and cluster IoT devices based on their sequential behavior patterns on enterprise network. In addition, due to limited resource and loose policies, IoT devices are more prone to be compromised than computers and servers in enterprise networks. Jisheng demonstrates a new IoT device behavior anomaly detection solution using a convolutional neural network (CNN) with an innovative way to encode device behavior into multichannel images.

Jisheng concludes by discussing the special challenges of applying deep learning into SaaS-based enterprise applications (e.g., using incremental learning to reduce the cost and dependency of retraining a whole neural network with all training data in private cloud deployments and using human feedback-based reinforcement learning to integrate human intelligence into the learning process to accelerate model convergence and adaption).

Photo of Jisheng Wang

Jisheng Wang

Aruba, a Hewlett Packard Enterprise Company

Jisheng Wang is senior director of data science at Aruba, a Hewlett Packard Enterprise company, where he leads the overall effort to apply data science in different enterprise network areas. Jisheng has over 12 years of extensive research and professional experience in applying state-of-art big data and data science technologies to solve challenging security problems. Previously, he was the chief scientist at Niara, where he led the overall innovation and development effort in big data infrastructure and data science, and was a technical lead for various security products at Cisco. Jisheng has published articles in top-tier publications and holds a number of patents, including one for an industry-first modular and data-agonistic UEBA solution. He holds a PhD in electrical engineering from Penn State University and an MS and BS in electrical engineering from Shanghai Jiao Tong University.