CODIN participated as a sponsor at the Persistent Homology Summer School in Rabat. During the conference, CODIN researchers Simona Giovannetti and Roberto Leuzzi spoke on the topic "Machine learning and Persistent Homology in Real Time Behavior Analysis".
Persistent Homology Summer school
Summary of the intervention.
Title: Machine learning and Persistent Homology in Real Time Behavior Analysis
Technology is constantly evolving offering new techniques to efficiently deal with the analysis of large amount of data often produced in real time. Data processing requires large amount of calculation time due to the intrinsic complexity of adopted algorithms and constant data growth. The rate of information growth together with the time it takes to process it, can make the results already obsolete while they are produced.
An effective behavior analysis solution must be able to efficiently process the data in streaming and in near real-time so as to ensure:
- consistency of the results with the analyzed situations;
- possibility of adopting a proactive approach;
- ability to actively respond to threats.
To support such solutions, the use of algebraic and machine learning algorithms is required. The technical and architectural implementation of these algorithms and related information structures becomes crucial to the success.
Two stage analysis, melding algebraic and machine learning approaches, has shown interesting positive results in performed bench marks to identify common and divergent behaviors.
Identified divergences from common behavioral patterns may indicate potential illicit, security risks or threats.