We invite the research community to submit innovative projects to collaborate on network development, including their applications, services and products.


RNP and Microsoft Challenge in Artificial Intelligence


The notice

In partnership with Microsoft, RNP launches the Artificial Intelligence (AI) R&D Challenge.

The purpose is to explore technologies, in addition to generating knowledge on the subjects proposed by the RNP technical committees. Microsoft will provide AI-related technology tools for the project development, offer specialized technical support and the documentation of these tools.

By means of this public notice, RNP invites the information and communication technologies (ICTs) research community to collaborate to this development.

Up to two projects were selected in each subject axis:

•    Network monitoring
•    Identity Management
•    Video collaboration

Closing of the proposal submission: 03/01/2019
Result release: 03/15/2019

Access the complete notice.

Proposal model .docx .odt



Selected Projects

Identity Management

eduroamIA: Artificial Intelligence Applied to Federated Authentication Event Prediction in eduroam - coordinated by Edelberto Silva (UFJF)
This project aims to predict critical events in the eduroam service by alerting its administrator at federation level, as well as at the institutional level, about anomalies related to the users´ authentication. Using artificial intelligence techniques and relying on the analysis of large volumes of service record data, both offline and online. Machine learning algorithms are applied to identify problems related to authentication of users of specific institutions and possible abnormal behavior correlated to the federated authentication. To achieve this goal, the RADIUS service records of users in roaming at eduroam federation level are used as input.

SAMD4IoT - Multilayer Authentication Service for IoT Devices - coordination by Kleber Cardoso (UFG)
Traditionally, device authentication has been based on use of encryption and sophisticated protocols. However, this approach has cost and complexity, which are incompatible with many types of IoT devices. This makes many IoT devices vulnerable to identity attacks, especially through wireless communication. To mitigate this problem, we propose an authentication service external to the IoT devices, which uses information from different layers. In the physical layer, the signature is identified by the peculiarities of each radiofrequency signal transmitter. In the application (or service) layer, the subscription is defined by a characteristic behavior of the service.

Network monitoring

Network Borescope: A tool for visual, intelligent, interactive and real-time analysis of backbone traffic - coordinated by Antonio Rocha (IC/UFF - Brazil)
Network performance monitoring is an extremely relevant subject for the scientific community and the industry in general. More recently, researchers and companies have been making efforts to use AI algorithms to solve problems related to the mentioned subjects. However, with the large volume of data and the need of efficient and interactive tools, this project proposes creation of a prototype tool, interactively and of efficient data structure, which uses ML.NET framework AI algorithms to identify patterns and predictions from geo-temporal data coming from real-time router monitoring.

Detection of Anomalies in Intelligent Environments Using AI - coordinated by Rafael Lopes Gomes (Universidade Estadual do Ceará (UECE) - Brazil)
The current smart environments are composed of IoT devices and end users (notebooks, tablets, etc.) communicating with each other and the Internet. The IoT devices are subject to anomalous behavior (non-standard operation) due to security vulnerabilities or malfunctions. Monitoring the behavior of these devices becomes crucial to assure efficient network performance. In this context, this project aims to develop a system for traffic monitoring and anomaly detection in intelligent environments supported by Artificial Intelligence (AI), generating a network profile and detecting possible anomalies through non-predicted traffic behavior.

Video collaboration

Research about Artificial Intelligence Algorithms to Assist in Diagnosis of Cataract by Telemedicine System - coordinated by Ronaldo Husemann (UFRGS)
The Teleoftalmo -UFRGS makes performs ophthalmologic exams on patients from the Unified Health System (SUS) through its own video collaboration tool, which enables real-time interaction between specialist doctors and patients in remote centers. Using this tool, the specialist doctor can talk to the patients and perform remote eye exams. However, the growing demand requires solutions that speed the issuance of reports up. This project proposes the development of an auxiliary Artificial Intelligence application, which can, by processing ophthalmic images, obtain automatic classification between Normal and Cataract patients, speeding the remote doctor´s decision-making up.

Detection of inappropriate content in video scenes - coordinated by Alan Guedes (Puc-Rio) and Sérgio Colcher (PUC-Rio)
The popularization of equipment to capture video and services for its storage and transmission has enabled production of a massive volume of video data. YouTube, for example, recorded upload of 72 hours of video per minute in 2014. While in 2018, that number increased to 400 hours of video per minute. This scenario presents a challenge to control the type of content loaded to these video storage services. For example, services, such as video@RNP (link sends email), ICD, and ITVRP constitute video sharing networks that focus on educational content with restrictions on inappropriate content. The classification of this type of content requires automatic analysis of this volume efficiently and practically. Methods based on Deep Learning (DL) have become state-of-art in different segments related to automated media analysis. This project focuses on assessing and developing such DL methods for detection of inappropriate content in video scenes.