SPATIAL takes its orientation towards state-of-the-art concerns and practices and then targets practical use cases concerning AI in ICT systems and cybersecurity.

Those are chosen to connect and promote business cases that are beneficial for the EU industry, including SMEs


What is our goal
Designing and implementing a Communications Framework for AI Privacy applications

What challenges do we face?
The performance of privacy-preserving machine learning – AI methods (classical, as well as deep learning methods) when deployed on edge computing nodes expected to be found in telco environments.

Desired Impact
The deployment of AI applications in Network virtualized infrastructure in an accountable and transparent way. Getting accuracy of AI algorithms while preserving users’ privacy.

Partner in charge
Telefónica

How SPATIAL will be utilized
Telefónica, within SPATIAL, will utilize a state-of-art piloting environment called 5TONIC, an Open 5G Lab recognized as a Digital Innovation Hub (DIH) by the European Union. The site already has a deployed network infrastructure for supporting pre-5G trials and a number of use cases.


What is our goal
To render the techniques used or introduced more transparent, resilient to adversarial ML, precise (e.g. by combining several ML algorithms) and efficient (e.g. using distributed data and processing of ML techniques).

What challenges do we face?
Cybersecurity risks, and protection of 5G and IoT networks, analysis of encrypted traffic, Root Cause Analysis.

Desired Impact
Better explainability, resiliency and distribution of AI/ML techniques. SPATIAL will allow improving the use and explainability of AI/ML techniques.

Partner in charge
Montimage

How SPATIAL will be utilized
The results of SPATIAL will be of great importance to improve these techniques, particularly concerning the use, resiliency, explainability and assessment of AI/ML. Techniques, guidelines and building blocks developed by the SPATIAL project will be considered and evaluated with respect to the improvements in the performance, transparency and precision they bring to the security analysis and algorithms.


What is our goal
The design and implementation of an explainable and accountable eCall functionality based on AI/ML algorithms within a Next Generation IP-based telecom platform with the capability of integrating rich-media emergency calls – a combination of voice, text, IoT sensors, social network information and video.

What challenges do we face?
Effective and accountable algorithms for analysing various data and automatically triggering an NG 112 call (i.e. eCall) in case of an emergency.

Desired Impact
Get coordination and communication among citizens, call centers and first responders in emergency situations.

Partner in charge
Fraunhofer Fokus

How SPATIAL will be utilized
AI/ML algorithms can be used within the available EMYNOS eCall demonstrator and must be made explainable and accountable for future product certification. In addition, the eCall demonstrator offers a high degree of complexity, so that the developed methods are robust and versatile, and easy transfer to other case studies can be achieved.


What is our goal
To study evasion and poisoning attacks against and defences for ML models used in cybersecurity.

What challenges do we face?
Privacy, data handling costs and security risks.

Desired Impact
Detection of advanced cyberattacks.

Partner in charge
F-Secure

How SPATIAL will be utilized
Implementing prototypes of dynamic attack detection systems (essentially data collection and analysis mechanisms) and experiment with attack tactics and techniques, to study their extent and key risks, and to validate and evaluate countermeasures proposed in SPATIAL.