AI Trustworthiness remains one of the major challenges from a tech and business perspective. Despite the fact that REWIRE is not fundamentally connected to AI regarding the notion of trust, the key values for a system operation are still commonly envisioned.
One of the key novelties of REWIRE is the integration of the AI-based training and detection process in the smart contract chaincode. More specifically, REWIRE project leverages FL as an AI-based enhancement on data filtering, vulnerabilities scanning, and execution monitoring so as to strengthen the security aspects of the overall process; particularly, focusing on device behavioural properties that cannot be captured by the designed cryptographic schemes and TEE-based abstractions. To this end, any deviations in the execution of the deployed SW, in the target device, will facilitate novel control-flow attestation methods based on the use of federated learning that enable us to offload all the monitored traces on the chaincode in order to perform in real-time.
Given the scope and utilization of AI, REWIRE’s ethics requirements, follow the principle of Responsible Innovation, on the basis mainly of the European Charter of Fundamental Rights, the EU guidelines on AI ethics, and the ethics guidelines for trustworthy AI. For Trustworthy AI, ALTAI represents an applied tool to support organisations and enterprises to self-assess their AI systems’ trustworthiness, following the Ethics Guidelines for Trustworthy Artificial Intelligence (AI). The ALTAI assessment tool is based on seven (7) pillars as requirements, which also describe the core requirements and ingredients for Trust. The seven pillars are the following:
- Human Agency and Oversight;
- Technical Robustness and Safety;
- Privacy and Data Governance;
- Transparency;
- Diversity, Non-discrimination and Fairness;
- Environmental and Societal well-being; and
- Accountability
In general, all these seven key areas must be adapted to REWIRE to ensure applicability. However, as we move from key area 1 out to 7, the values become more generalised, and hence, the adaption needs to become more general. That is, the first two key areas are much more easily translatable to specific engineering requirements and existing standards, whereas the other five areas apply to the Compute Continuum more generally and must also be understood by reference to existing law, legislation, and policy, as well as emerging and evolving social norms.
To this end, within the REWIRE context, the consortium manages to directly incorporate the applicability of the key values from the technical standpoint, which are the following: (2) Technical robustness and safety; (3) Privacy and data governance; (4) Transparency; and (7) Accountability.
REWIRE provides state-of-the-art attestation mechanisms towards establishing robustness and safety for Compute Continuum systems, which rely on technical robustness to both ensure and assure users of their reliability and safety. This approach is essential to mitigate and prevent even unintentional harm effectively. In addition, in order to ensure full adherence to privacy and data protection standards, robust data governance mechanisms must be established. These should focus on maintaining data quality and integrity while providing legitimized and controlled access to the data. REWIRE’s approach supports zero-touch onboarding aligned with the SSI concept. Finally, for accountability, REWIRE capitalizes on Blockchain technology, where every transaction is linked through cryptographic hashes, allowing detailed and transparent audits, while for accountability, robust mechanisms must be established. REWIRE’s Blockchain technology through its immutability feature records all transactions and changes in a secure and accountable manner.