Our research goal is to move AI from associating predefined reasonings about data, and the concept of an unconditioned, universal and current truth, into a simpler representation of multimodal insights capturing the complexity of the world around us. The components of our research include:
- Perception: Expanding the current human perception with unexplored stimuli from new interfaces and techniques
- Interpretation: Define predictive and casual reasoning models on digital data capturing the complexity and dynamicity of the world around us
- Communication: Supporting verbal and non-verbal semantic and emotional communication
Our goal is to extract meaningful insights from IoT Data to augment data exploration. This research leverages machine learning, novel user experiences and interfaces to create an automated data science exploration system. We explore new ways of representing information that humans can use.
Networked Systems and Automation
We are applying machine learning and other algorithmic approaches to helping move today’s human-centric networks to a machine-centric networks filled with sensors, bots, robots, drones and smart objects where every object is network attached. With no bound to the number of machines, hyper-scaling to billions of nodes will be required to keep machines connected and functioning.
Network automation and optimization
Challenges associated with hyper-scaled, low latency, machine-oriented networks includes advanced routing algorithms for SD-WANs, MPLS, Segment Routing optimizations, modular data planes and data plane acceleration for SDNs, Optimizing networks for IoT and Industrial automation, performance characterization of new data center switch architectures.
Networks that learn
Networks will need to automate their operations far beyond the current state-of-the-art to support the envisaged level of scale efficiently. Networks will need automated techniques to configure and onboard devices, secure the network from the vast attack surface posed by the large number of connected devices, optimize the network to carry more revenue bearing traffic while providing an enhanced Quality-of-Experience (QoE) to endpoints, and predict future demand, network hotspots and equipment outages. For hyper-scale automation, the network needs to self-adapt to different requirements automatically with virtually no human intervention.
Specific Research Projects
- Fingerprinting Microservices
Machine learning identifies containerized microservices from observations external to the container such as the sequence of system calls that the containerized application makes. The algorithm scales and identifies the broad functionality of the microservice and the software platform used to implement the application.
- Traffic Fingerprinting for Network Attack Detection
Intrusions in industrial IoT networks are identified, and SDN-based threat mitigation restores operations. The challenges are in developing machine learning methods for “fingerprinting” normal traffic in these networks, based on monitoring traffic over time at various vantage points in the network, so that anomalies are quickly identified.
- Accelerating Blockchains
Energy-intensive blockchain computations are simplified with “singular block”inspection to reduce transaction confirmation latencies to milliseconds. We are looking at ways to remain compatible with existing implementations while enabling a wide range of applications.
- Performance of Partially-Buffered Switch Architectures
Conventional fast and deep buffering used in routers can evolve to partially buffered switches having a variety of buffer speeds that adapt to the overall traffic needs. We study real-world traffic to develop new switching architectures with higher performance.
- Programmable Data Plane
We are implementing network virtualization that goes beyond the control plane. With our ClickNF implementation, we provide libraries of modular transport and application-layer building blocks for the development of middleboxes and server-side network functions. ClickNF is open source and was made publicly available on Nokia’s GitHub page in 2017.
Reddit and Twitter are powerful means of relevant content discovery, not because of their recommendation algorithms, but because large user groups contribute their opinions. WeBrowse helps communities find popular content and relevant information without the need for active user engagement, by looking at more subtle cues of user interest.
If you would like to learn more about research at Bell Labs, enter your return email address here and describe your specific interest. Our staff will forward your message to the appropriate party.