On the use of similarity search to detect fake scientific papers