Results and Impact
The main result from the AIDA project is, therefore, the prototype of what Mobileum envisages as the next generation of the RAID platform.
A complete prototype of AIDA will be deployed and demoed by the end of the project in a realistic scenario in the sector of the telecommunication services.
RAID’s platform is currently applied around the world in revenue assurance, business assurance and fraud management, among others.
It also opens market opportunities, namely through the exploration of emergent federated machine learning techniques that will reinforce the analytic capabilities of the platform and may, in certain cases, obviate confidentiality issues.
p-sp Solution
What is it? A distributed database system made to be geo-scaled, to serve customers across the globe with low latencies. This system was developed by INESC TEC.
Example: p-sp allows strong guarantees of consistency, which is ideal for applications such as “Amazon” that need to sell products to multiple points around the globe.
Polyglot Solution
What is it? The polyglot service essentially allows access to data stored in different databases. Databases exist for different purposes, and this service allows, using the same language (sql), to query any database.
Impact: By having a polyglot service, we can easily join data from different databases in real time. The alternative was to first load everything into the same database and only then do the data processing, which is more costly and time consuming.
Example: On Edge, it is common to have multiple databases – a MongoDB to store documents, a Cassandra to store temporal data, etc. With the Polyglot solution, we have a central decision application in the cloud that relates the data in MongoDB to the data collected in Cassandra, without needing to change the way the data is stored.
IRSF Solution
What is it? International Revenue Share Fraud (IRSF) is one of the project’s use cases – Fraud Detection using Machine Learning. The IRSF solution allows to control all traffic CDRs generated at the network switches combined with the use of near real-time exchange of records on both voice and data, providing the detailed data required to detect complex and elusive fraud patterns quickly, allowing to reduce fraud detection time and losses typically by 90% or more.
Impact: The main impact is to reduce a company’s loss of money due to a variety of frauds. This particular prototype is for cell phone fraud, but AIDA has more, such as detection of illegal use of streaming services.
