Online sales are at an all-time high. More than every second euro is spent on large online marketplaces. In parallel, there has been further increase in cybercrime offenses. Two issues are hereby of neuralgic importance for consumers: fraud by fake shops and fraud by fake investment platforms, which is increasing in line with the popularity of cryptocurrencies. Fake shops cause great economic harm; a dark field study assumes 320,000 directly affected consumers in Austria and estimates the amount of damage at 16 million euros. 

The KIRAS project SINBAD has successfully developed a fake shop detector based on artificial intelligence (AI). This is available for consumers to download free of charge as a browser plugin for Edge, Firefox and Chrome. The AI models in place achieve an accuracy of 91% in practical use proven on over 400,000 websites. The trained AI system has over 21,000 features available for decision-making. The strength of the method lies in the fact that no individual feature stands out, but that the combination of a large number of individual characteristics, including their presence or non-existence, leads to a very robust risk assessment by the AI.

The project "Resilience in Online Trade" (RIO) continues the successful preventive work through targeted innovations along the fake shop detection lifecycle. These include:

1. A modular, scalable, and easily expandable open-source platform for AI-based risk assessment services and their applications for quality-assured practical use.
2. A community-enabled fraud prevention approach: Through the use of AI detection, the number of published warnings reached a new high. It is important to support and relieve experts by delegating quality assurance tasks to the community, achieved via suitable gamification and nudging approaches in the form of an online game.
3. The implementation of a minimally invasive app-based solution for real-time protection against fraudulent e-commerce increases the mobile resilience while maintaining high privacy standards.
4. The development of demonstrators with a focus on Natural Language Processing (NLP) aim to increase the human explainability of AI-based risk assessments, detect related fraudulent instances (clusters) and implement fraud prevention measure on large online marketplaces. This is done with the support of the BMSGPK and BMI and includes the evaluation of the demonstrated potential in the stakeholder’s context with regard to supplementary effects of existing preventative and investigative measures.
5. Porting and applying successfully used tools and methods from the fake shop detection scenario to the domain of fraudulent cryptocurrency investment platforms to build up protective measures
   against this growing threat to consumers.
6. Building up knowledge for targeted preventative measures, via two studies on "sociodemographic factors of AI-based trust-calibration" and a "dark field study of those affected, exposing fraud patterns and gray areas in online trade”

RIO is funded by the KIRAS security research funding programme of the Federal Ministry of Finance.