COMSNETS 2019

Programmable networks can offer very efficient and granular control of resources to applications and users. Consider a situation where 100s of users belonging to a certain project can be automatically be onboarded (assigned the right IP subnet, policies, access) on the network using Rest APIs. Or another situation where based on the user activity, the user can be quarantined in the network itself rather than at the application. All of these can be achieved by intelligent apps which will use REST APIs exposed by a network controller. The demo/exhibit we propose will showcase DNAC (Digital Network Architecture Controller), the REST APIs it exposes and also apps that have been built using these APIs.

The DNAC is a centralized configuration, management and analytic engine that manages the whole network. It can program policies on the network, segment the network for different users for security, analyse the data of the network and make intelligent decisions and give a single point of control to the network. What makes it even more amazing is the REST APIs it exposes that will enable apps to be written that can extract the potential of the network to its fullest.

TensorFlow is a machine learning framework, with a focus on neural networks. We have enabled the TensorFlow runtime engine, known as TensorFlow Serving, to run on Cisco IOx, thereby enabling IOx Fog applications to easily use neural network models on large amounts of data available at the edge, to derive local intelligence efficiently and quickly. In this demo, we showcase a demo IOx application that receives camera feed from a local camera, to detect human figures by using a TensorFlow model trained on human figures, and subsequently perform selective reporting to a cloud dashboard, thereby avoiding bandwidth costs of transferring the entire camera video stream to the cloud.

Come and immerse yourself into rich sensory experiences on Qualcomm’s mobile XR devices. The Snapdragon XR platforms are engineered to support high-quality UltraHD premium video with fast graphics and low power consumption, allowing users to lean back and immerse themselves in their favorite movie, videos, and sports.

The proliferation of Internet of Things (IoT) devices has led to the deployment of various types of sensors in the homes, offices, buildings, lawns, cities, and even in agricultural farms. Due to the diverse nature of IoT deployments and the likelihood of sensor failures in-the-wild, a key challenge in the design of IoT systems is ensuring the integrity, accuracy, and fidelity of sensor data.

We present a system based on the Fall-curve primitive -- a sensor's voltage response when the power is turned off -- to characterize the sensor. A sensor's Fall-curve constitutes a unique signature using which it is possible to identify the sensor and also detect whether it is operating correctly or not. In this demo, we show Fall-curve in action to accurately detect and identify the sensors connected to an IoT device. Furthermore, we also show that Fall-curves can reliably detect various transient and permanent sensor faults in an IoT device.

In this demo, we will go over basic components of Trusted AI. We will show how Data Bias detection and mitigation algorithm works in practice. We will argue how end to end provenance can help add extra level of trust with the consumer. Finally, we will go over IBM's hybrid approach to handle this complex problem. Business users can make sure of this technology via interactive interfaces where core AI developers can start by using our opensource toolkit AIF60.

NetSim is a leading network simulation and emulation software for protocol modeling and simulation. It enables modeling, simulation and analysis of a wide range of wired, wireless, mobile and sensor networks. NetSim is being used by over 400+ customers across 20 countries.

The paper outlines the features, use cases, modules and key success stories of NuSpace - a multi-user, multi-device, multi-location collaboration platform in Digital Reality. NuSpace is the flagship product of Imaginate, a company providing exciting products and services in the area of Virtual,Augmented, and Mixed reality. NuSpace provides end-to-end infrastructure for geographically distributed teams to collaborate over contextual 2D/3D data. This makes it a great platform to fully tap into the vast powers of immersive realities to improve process efficiencies, enhance employee productivity and reduce travel expenses. The platform is being used across a wide range of industries for collaborative training, design reviews, support and maintenance, guided walk throughs and remote sales. Imaginate would like to demonstrate the power of collaborating in an immersive world during the COMSNETS 2019 conference.

Ostinato is an open-source packet generator and network test tool that aims to provide a traffic generator to every network engineer, developer and tester to help improve their productivity, effectiveness and the overall product quality.

It hopes to replace the highly expensive commercial solutions from companies like Ixia, Spirent etc. which are a scarce resource due to their prohibitive cost even for major networking vendors like Cisco and Juniper.

Ostinato is of interest to all COMSNETS attendees including industry folks, teachers, students and researchers. And Ostinato will benefit by this exposure and marketing opportunity.

Large-scale WiFi networks are becoming increasingly common in telco and distributed enterprise scenarios. Such networks comprise 1000s of Access Points, deployed over 100s of locations. While the management of each location is typically done by teams at the local level, there is generally a centralized Network Operations Center (NOC) which manages the entire network. NOC personnel require both a bird's eye view of the entire network as well as in-depth details of each location. Hence, there is a need for a dashboard that can provide insights into the network performance at multiple levels of spatio-temporal granularity. We have built such a dashboard using a set of Key Performance Indicators (KPIs), periodically generated by the network, stored in CSV files. A Big Data platform has been setup which pulls these files and abstracts the KPIs as time and location-based visualizations, providing insights into network performance across different metrics.

With the increase in foot print for networking services using cloud based technologies, network service providers/manufacturers have to effectively test the cloud services with limited resources and time in cost effective manner. The Network-in-a-box offers to quickly deploy the cloud networking services and validate the solutions needed by the field.

The Juniper NFX devices offer deployment of Secure SDWAN with service chaining using Juniper and other 3rd party VNFs integrated in a single box. This provides an excellent platform to test, validate and automate the customer deployment use cases. The VNFs that can be supported by NFX includes vSRX(SDSN) and various 3rd party VNFs. Using the REST APIs end-to-end interactions can be automated which enables the vendors to quickly validate and regress the changes.

The demonstration includes creating 5 VNFs in Juniper NFX - 1xtraffic generator, 1xL2/L3 service chain, 2xVSRX and 1xCentOS(to simulate network impediments).

Lower-body biomechanical analysis is crucial for diagnosing and rehabilitating certain injuries and ailments. Conventionally, this has been achieved using video cameras and motion sensor based systems. However, these systems are limited in usability to in-clinic only, and are cumbersome to set up and drive. We present a smart insole based system: Stridalyzer INSIGHT, consisting of insoles embedded with 6-axis motion sensor (accelerometer & gyroscope) and pressure sensors, which connects via Bluetooth to its mobile application with patent-pending real-time and context-aware analytics and cloud-based engine for long-term analysis. The device finds its application in prevention and diagnosis of injuries, rehabilitation and biomechanical research. The insole sensors provide real-time pressure maps for key pressure points, as well as linear & rotational motion of the foot. Using these data, the application performs accurate analysis of stride properties like center of balance (COB), ground contact time (GCT), degree of pronation, peak and mean ground reaction forces (GRFs). Long term trend analysis for specific gait parameters for each user/patient is provided by cloud software.

Videos are hard to peer inside. The rich information hiding inside the videos needs to be uncovered. All textbooks come with a table-of-contents and a glossary, which summarize the key topics being covered by the textbook and help the reader identify which pages cover a specific topic. Why should informational videos be any different? However, since millions of informational videos already exist without a table-of-contents or glossary, there is a need to create those indices automatically.

At a high level, the key problem is one of user engagement for informational videos, and there are related problems of topic discovery and search within videos. This is an open problem, affecting hundreds of millions of viewers of videos on the internet.

VideoKen is an AI-based platform that makes any informational videos richer and more consumable by automatically indexing videos, creating table-of-contents (ToC) and phrase cloud, which nicely summarizes key topics in a video. It works on cutting edge technologies to uncover rich information hiding inside the videos. Our novel video processing pipeline implements customized state-of-art AI techniques, which deal with multi-modal unstructured data, including text, images, audio and enables a powerful and new video viewing experience for our end users. VideoKen's embedded video player provides unique navigational capabilities to jump within a video to points of interest to the viewer. VideoKen further provides insights on which topics within a video received more views, and where the viewers left the video. We solve the problem of "opaqueness" of videos, due to which most viewers don't watch past the first few minutes of an informational video. Our data shows viewer engagement is increased 2x or more with our indexing features.

Here's a quick overview of the product: https://videoken.com/embed/iCkM4eccQAQ

You can experience the technology at https://videoken.com

We developed a smartphone-based retinal camera with an on-device offline low latency diagnosis capability based on deep learning. This opens doors for better healthcare in remote rural regions.