Periodic Reporting for period 1 - ROBUST-6G (SmaRt, AutOmated, and ReliaBle SecUrity Service PlaTform for 6G)
Berichtszeitraum: 2024-01-01 bis 2024-12-31
To meet the complex demands of 6G use cases, 6G will require full automation of network and service management, while mitigating cybersecurity risks posed by its expanded threat landscape. The ROBUST-6G project is focused on developing an AI/ML-driven, zero-touch, end-to-end 6G security framework. ROBUST-6G prioritizes the security of distributed intelligence and promotes transparent, green, and sustainable AI/ML solutions to minimize energy consumption and foster trustworthy 6G networks. ROBUST-6G places a special emphasis on Physical Layer Security (PLS) as the first line of defense for wireless networks. ROBUST-6G aims to develop state-of-the-art techniques for anomaly detection, attack classification, and device authentication at the physical layer, creating a secure and efficient 6G ecosystem.
ROBUST-6G's main objectives:
1. Analysis of anticipated 6G architecture and scenarios, identification, and characterization of the threat landscape in an AI-driven 6G Networks.
2. To develop a holistic E2E 6G security architecture with inherent AI functionalities that will seamlessly integrate different functions in a heterogenous network environment.
3. Development of robust, sustainable (in terms of energy efficiency), explainable, effective (in terms of performance) and preserving privacy AI-driven security functionalities.
4. Automatic, zero-touch, security, and resource management for trusted and certified services among multiple stakeholders in distributed dynamic scenarios.
5. AI/ML-enabled smart techniques to detect and mitigate physical layer attacks on network and user devices and to propose novel physical layer security schemes for demanding scenarios, taking into account new radio technologies for 6G.
6. Validate the ROBUST-6G innovations through three use case scenarios.
ROBUST-6G's main objectives:
1. Analysis of anticipated 6G architecture and scenarios, identification, and characterization of the threat landscape in an AI-driven 6G Networks.
2. To develop a holistic E2E 6G security architecture with inherent AI functionalities that will seamlessly integrate different functions in a heterogenous network environment.
3. Development of robust, sustainable (in terms of energy efficiency), explainable, effective (in terms of performance) and preserving privacy AI-driven security functionalities.
4. Automatic, zero-touch, security, and resource management for trusted and certified services among multiple stakeholders in distributed dynamic scenarios.
5. AI/ML-enabled smart techniques to detect and mitigate physical layer attacks on network and user devices and to propose novel physical layer security schemes for demanding scenarios, taking into account new radio technologies for 6G.
6. Validate the ROBUST-6G innovations through three use case scenarios.
* A comprehensive review of existing solutions for threat detection and protection, alongside an in-depth analysis of security threat characteristics in future 6G networks. The analysis focuses on key 6G technical enablers, use cases, and applications, with an emphasis on physical layer threats, AI/ML modules, and Application Programming Interfaces (APIs).
* The initial threat analysis in D2.1 has been expanded to include specific prevention and mitigation mechanisms for threats targeting AI/ML, detailed in Deliverable 3.1 Threat Assessment and Prevention Report.
* For the identified physical layer threats in D2.1 relevant datasets have been reviewed and mapped to the corresponding threats for future research activities in D5.1.
* Generated a consolidated high-level architecture of ROBUST-6G where principal functionalities, services, and components are depicted. The high-level architecture offers a summary of the technologies planned for development in the project and outlines the system’s goal of achieving comprehensive AI-driven security in the envisioned 6G networks. This study has been documented in Deliverable 2.2 Use Cases, Requirements, ROBUST-6G Initial Architecture and Initial ROBUST-6G Dataspace.
* A critical component for managing and processing the substantial data volumes generated in a 6G environment, Data Fabric has been incorporated with advanced security mechanisms, such as fine-grained access control using tools like Open Policy Agent (OPA) and Keycloak.
* A decentralized federated learning framework has been developed for privacy-preserving ML, with efforts focusing on robustness against adversarial attacks and implementing privacy-preserving measures such as homomorphic encryption.
* The functional architecture of the ROBUST-6G Zero-Touch Security Platform has been designed.
* A comprehensive validation framework, outlining methodologies such as scenario-based testing, KPI-driven validation, and incremental integration have been introduced Deliverable D6.1.
* Collected testbed data to support experimental designs and validation strategies, with an emphasis on establishing parameters for evaluating use case components.
* The initial threat analysis in D2.1 has been expanded to include specific prevention and mitigation mechanisms for threats targeting AI/ML, detailed in Deliverable 3.1 Threat Assessment and Prevention Report.
* For the identified physical layer threats in D2.1 relevant datasets have been reviewed and mapped to the corresponding threats for future research activities in D5.1.
* Generated a consolidated high-level architecture of ROBUST-6G where principal functionalities, services, and components are depicted. The high-level architecture offers a summary of the technologies planned for development in the project and outlines the system’s goal of achieving comprehensive AI-driven security in the envisioned 6G networks. This study has been documented in Deliverable 2.2 Use Cases, Requirements, ROBUST-6G Initial Architecture and Initial ROBUST-6G Dataspace.
* A critical component for managing and processing the substantial data volumes generated in a 6G environment, Data Fabric has been incorporated with advanced security mechanisms, such as fine-grained access control using tools like Open Policy Agent (OPA) and Keycloak.
* A decentralized federated learning framework has been developed for privacy-preserving ML, with efforts focusing on robustness against adversarial attacks and implementing privacy-preserving measures such as homomorphic encryption.
* The functional architecture of the ROBUST-6G Zero-Touch Security Platform has been designed.
* A comprehensive validation framework, outlining methodologies such as scenario-based testing, KPI-driven validation, and incremental integration have been introduced Deliverable D6.1.
* Collected testbed data to support experimental designs and validation strategies, with an emphasis on establishing parameters for evaluating use case components.
* A decentralized federated learning framework has been developed for privacy-preserving ML, with efforts focusing on robustness against adversarial attacks and implementing privacy-preserving measures such as homomorphic encryption.
* A novel data poisoning and inference attacks on the fully DFL systems with the use of the Layerwise Relevance Propagation (LRP) technique.
* DRACO, a novel method for decentralized asynchronous Stochastic Gradient Descent (SGD) over row-stochastic gossip wireless networks by leveraging continuous communication.
* A peer-to-peer federated learning python-based program where homomorphic encryption is integrated to protect model updates against privacy attacks.
* A novel method for characterizing and enhancing the trustworthiness of a prediction in anomaly detection in IDS before its evaluation.
* A regularization technique for XAI, leveraging SHIELD (Selective Hidden Input Evaluation for Learning Dynamics), a regularization family designed to improve model quality
* A Security Orchestrator (Z-SO) that integrates with various orchestration levels to manage AI/ML-enabled security services in a cloud and edge environment.
* A semantics-aware approach to remote estimation of Markov sources by integrating the Age of Missed Alarm (AoMA) and Age of False Alarm (AoFA) metrics
* An adaptive threat mitigation method at the PHY, incorporating AoA as an inout feature in a trust and reputation management (TRM) module to dynamically adapt trustworthiness metrics.
* A new metric, Age of Consecutive Error (AoCE), to prioritize the timely detection and resolution of critical errors
* A novel challenge-response PLA (CR-PLA) mechanism for a cellular system that leverages the reconfigurability property of a reconfigurable intelligent surface (RIS) in an authentication mechanism.
* A new attack against challenge response physical layer authentication (CR-PLA) with intelligent RISs.
* A cross-layer trust and reputation management framework in 6G has been proposed.
* A novel data poisoning and inference attacks on the fully DFL systems with the use of the Layerwise Relevance Propagation (LRP) technique.
* DRACO, a novel method for decentralized asynchronous Stochastic Gradient Descent (SGD) over row-stochastic gossip wireless networks by leveraging continuous communication.
* A peer-to-peer federated learning python-based program where homomorphic encryption is integrated to protect model updates against privacy attacks.
* A novel method for characterizing and enhancing the trustworthiness of a prediction in anomaly detection in IDS before its evaluation.
* A regularization technique for XAI, leveraging SHIELD (Selective Hidden Input Evaluation for Learning Dynamics), a regularization family designed to improve model quality
* A Security Orchestrator (Z-SO) that integrates with various orchestration levels to manage AI/ML-enabled security services in a cloud and edge environment.
* A semantics-aware approach to remote estimation of Markov sources by integrating the Age of Missed Alarm (AoMA) and Age of False Alarm (AoFA) metrics
* An adaptive threat mitigation method at the PHY, incorporating AoA as an inout feature in a trust and reputation management (TRM) module to dynamically adapt trustworthiness metrics.
* A new metric, Age of Consecutive Error (AoCE), to prioritize the timely detection and resolution of critical errors
* A novel challenge-response PLA (CR-PLA) mechanism for a cellular system that leverages the reconfigurability property of a reconfigurable intelligent surface (RIS) in an authentication mechanism.
* A new attack against challenge response physical layer authentication (CR-PLA) with intelligent RISs.
* A cross-layer trust and reputation management framework in 6G has been proposed.