Reliable AI / Machine Learning for IoT
InSecTT will considerably strengthen Europe’s position on fully utilising AI and machine learning in Cyber IoT Systems. The data gathering and complexity of the IoT systems is increasing rapidly and new methods are required to manage this growth. AI methods provide significant benefits when industry is targeting on more intelligent and better performing CPS products, but at the moment the AI development and verification models as well statistical behahiour of the neural networks prevent large scale utilisation of the machine learning technology, due to black box nature that sometimes may lead into surprising outcomes. AI will have significant impact on economy as technology itself is providing large growth and it will also change operating models on many traditional fields.
AI based economic growth is expected to have 45% annual growth and US and China are leading the transition. Europe has potential to catch-up, but it requires substantially more investments to keep Europe competitive in global markets. Source: Accenture Frontier Economics.
AI technology is changing how IoT systems provide intelligence, autonomy, security and self-monitoring capabilities. InSecTT results will provide participating companies new methods, tools and technologies to apply ML methods in their commercial offering. This will increase the competence of the participating companies in general for developing AI solutions as well as increase the competitiveness of their products in the market. As InSecTT consortium is well connected with European industry the project results will increase European capabilities and know-how for catching up the expertise gap between Europe and other markets, such as China and USA.
A study by Accenture and Frontier Economics expects a big impact of AI on a contries gross value added (GVA). They estimate the annual GVA growth rates in 2035 in a baseline state with current assumptions on expected economic growth and a AI steady state, assuming Artificial Intelligence being integrated into economic processes.
The impact ranges depending on the country from 0.8 percentage points increase in potential GVA growth rates in Italy or Spain to 2.0 percentage points in Finland or the U.S. (Source: Accenture Frontier Economics)
Machine learning activities in InSecTT will lower the barriers of developing next generation AI powered IoT systems and provide competitive edge for European industry, leading into increased revenues, safety and customer experience as well as reduced costs.
Safety and trust
InSecTT AI development contribute to the safety and reliability of the IoT systems by providing methods for explaining the decision making as well as increase systematic approach for testing and validating AI technology. In addition, InSecTT will create a trustworthiness framework that leads faster into solutions that are accepted and trusted by the end users and other stakeholders. More robust and reliable AI models provide more predictable behaviour for IoT systems as well as increase the robustness and cyber security of the wireless communication interfaces. When autonomy of the devices is increased the ethical aspects in the decision-making process needs to be taken into account as well. The explainable model reveals the reasoning behind decision making that allows inspecting the behaviour on different aspects, such as ethics, fairness and privacy.
Performance of the IoT systems can be measured in multiple aspects. Different systems need to have various characteristics related to the robustness, power consumption, latency, security, data quality, complexity, costs of set-up, and maintenance, etc. The AI development allows optimizing the system performance towards preferred targets, making solutions more dynamic and competitive on markets. InSecTT will develop solutions for increasing the trust, wireless performance, security, distributed computation and validation giving a competitive edge for participating companies and Europe in general.
Applying ML in IoT system development requires new approaches for verifying correct operation in all conditions if possible. InSecTT will develop explainable AI methodology, security methods and validation tools that provide tools for verifying ML functionality and user trust in a systematic way. These results and increased knowhow enable faster time to market and better user acceptance for new AI powered commercial offering.
AI processing requires considerably more computing power than typical software solutions in nowadays IoT systems. One barrier for widely increased use of AI technology is the increased cost of the hardware resources and power consumption required to perform neural network processing. InSecTT will develop new HW for cost efficient implementation of ML processing on the device as well create models for distributing the computation on different layers of the network topology. InSecTT architecture consists of IoT devices, edge/fog computing nodes and cloud processing. The results will allow more flexible, cost-efficient and power saving implementation of ML computation facilitating the development of novel ML enhanced IoT solutions.