Based on the situation report of the Federal Ministry of the Interior, eCommerce fraud reached an all-time high in 2019 (+ 32.3%) - over 10,500 reports from consumers were received by the Austrian agency Watchlist Internet during this period including illegitimate subscription services, counterfeit goods and fake-shops. Factors such as the price or advertisements on social media play a decisive role in specifically addressing customers, and clever tricks such as customized landing-pages make enforcement of those affected difficult.

Prevention is a key instrument in fighting cybercrime. However, consumers hold back on reporting incidents, information reaches the experts in fragments and most often harm has already occurred before fake-shops are flagged. Technological solutions that actively warn and protect consumers are an important addition to preventive measures. Machine learning (ML) methods in this field have increasingly improved in the past five years. In Austria, the ongoing flagship project MAL2 enables the classification of fake shops with detection rates of over 90%. Data analysis presents an above-average degree of cluster formations - this leads to the hypothesis of the existence and usage of modular fake shop systems and partially organized crime; there are also indications of money laundering by fake shops. The limits of current research approaches to investigate these suspicions are defined by the use of shallow ML models due to insufficient data. The models lack robustness with regard to reliability and the small number of potentially available intrinsic features to explain the cluster formations, as well as their decrease in accuracy over time.

Based on this, SINBAD derives the specific need for security research in the following areas: exploring gaps in the information space; providing specific advancements; proactively protecting consumers and developing counter-narratives. A comprehensive analysis of fake shop fraud is carried out via (1) a technology-based monitoring with real persons for collecting data on listed products and prices, (2) a dark web research on modular fake shop systems, (3) analysis of ways and methods of group targeting by fraudulent eCommerce (advertising, social media, search engines) as well as assessing consumer needs. Building on existing MAL2 results, SINBAD is implementing a multi-task ML system for the detection of fake shops that autonomously decides when parameters must be taken into account in the decision-making process. By proactively screening newly registered domains in the DACH region, the efficiency of the fake shop AI detector prototypes in terms of its ability in decreasing the window of opportunity (WoO) of fraudulent offers is evaluated and an interdisciplinary catalog of measures is derived.

The goals of the project are to gain new insights into means of proactively detecting fake shops through user-centered methods, data-based models and the deepening of machine learning processes; to develop effective counter-narratives - under the involvement of the stakeholder BMASGPK which are based on the exploitation of key project results and through stakeholder dialogue with members of politics, consumer protection and e-commerce – through which consumers are strengthened and protected.

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