You want to detect fraudulent accounts and identify whether a newly created account matches any previously known fraudulent accounts. You are cleansing logs with AWS Glue during ingestion. Which approach should you use to find likely duplicate or matching fraudulent accounts?

The AWS Glue FindMatches ML transform is specifically designed for record linkage and deduplication tasks, even when exact matches don't exist. It uses machine learning to identify records that refer to the same real-world entity, which is ideal for finding likely duplicate or matching fraudulent accounts despite variations in data. Running a built-in FindDuplicates query in Amazon Athena is not a standard feature for fuzzy matching and would only find exact duplicates. An AWS Glue crawler infers schema and partitions, not duplicate accounts. Searching for duplicates directly in the AWS Glue Data Catalog is not a functional capability; the Data Catalog stores metadata, not a search engine for data content.