The Ubcoin Market Identifies Suspicious and Fraudulent Ads Utilizing Artificial Intelligence
Ubcoin is a unique crypto-to-goods exchange — a blockchain-based platform where everyone can buy or sell real goods for cryptocurrency. In marketplaces like Ubcoin Market listings (ads) are created and posted by users of the service. Among the legitimate ones some are posted by scammers. To search and identify listings posted by scammers, the Ubcoin Market team developed a solution based on artificial intelligence technologies. It identifies suspicious ads, reveals duplicate ads, low-cost ads, abnormal ads and finds weird user behavior patterns in the data. The use of machine learning algorithms makes the purchasing and selling of goods for cryptocurrency on the Ubcoin Market safe and secure.
Advertisements posted by scammers may barely differ from usual legitimate ads, which makes preliminary moderation using the involvement of human experts ineffective. The algorithms utilized by Ubcoin artificial intelligence solution finds obscure patterns of behavior of scammers and instantly identifies suspicious ads.
The Ubcoin team identifies four types of fraudulent ads:
1) Duplicate ads, with changed details or contacts
2) Ads, with a low price, for the mass attraction of buyers
3) Multiple ads placed by one user from different accounts
4) Ads containing abnormal content or abnormal user behavior when posted
To identify duplicate ads Ubcoin Market uses its own solution for the analysis of texts and images. Natural language processing algorithms such as Word2Vec translate text into a vector and define the same meaning even when using substitute words. A solution for image analysis reveals duplicate images and select objects on them.
To identify ads with a low price, a solution is used to automatically determine the optimal price for the product. The solution utilizes real historical data regarding the cost of placement and the sale of goods. The solution is based on the algorithm for highlighting the topic of the ad and the characteristics of the product specified in the ad. If the cost of the goods is understated and falls into the left 5% interval of the price distribution histogram, this is a sign for additional verification.
To detect multiple ads placed by one user from different accounts, both the data received when a user visits the site and the data from the ad are used. The algorithms for clustering are used to reveal the hidden structure in the data. The approach reveals obscure behavioral factors. An example of the operation of the clustering algorithm: clusters are marked with colors. When finding a cluster in which fraudulent ads predominate, a classification algorithm is trained based on the information received, which automatically detects suspicious ads in the future.
Additionally, a solution is used to search for anomalies in the data. The solution is based on the one-class algorithm of the method of support vectors. Search for anomalies is designed to search for ads that are very different in their own structure or the nature of the user’s behavior from the bulk of the ads. The solution finds ads posted by bots, and automatically sends them for additional verification.
The peculiarity of revealing fraudulent actions is that people who aspire to commit unlawful actions quickly adapt to new ways of counteraction, changing behavior. To solve this problem, artificial intelligence Ubcoin Market will undergo regular retraining on new data, and the analysts and developers will continuously improve solutions to identify suspicious transactions.
The decision of the Ubcoin Market to identify suspicious ads is based on modern machine learning algorithms and identifies duplicate ads, low-price ads, abnormal ads and user behavior patterns. The use of machine learning algorithms makes the purchase and sale of goods for cryptocurrency on the Ubcoin Market secure and safe.
No comments:
Post a Comment