Loopholes of Plagiarism Detection Software

AbstractNowadays, information becomes more and more important for our society. We can get information from everywhere easily and fast. The release of translation software eliminated the problems of translating a document. As a result of the continuous development, recent software are handy and sophisticated. This new way of getting information simplified stealing other people's spiritual work. Plagiarism becomes frequent. The number of students who submit other's work as their own is constantly increasing. The Internet gives them this opportunity. This growth won’t stop as far as we don’t try to detect plagiarism.The scandals of recent past made plagiarism detection a popular issue. The topic has chosen, because the further aim of this work is to develop a software, which meets our world's requirements. Throughout my work some software have been tested, in a few cases, those failed in finding plagiarism, so now it is well-known, what the loopholes of these systems are. The future work is supposed to fix these problems, making my software reliable in order to stop plagiarism among universities.The first tested software was the KOPI online plagiarism detector, which is developed and operated by MTA SZTAKI Department of Distributed Systems. The system compares documents with the ones found in its own database, which contains other users’ documents and articles from the English and the Hungarian Wikipedia.Plagiarism detection with KOPI has four steps. It chunks the document into 40-60 characters long fragments. Storing these fragments results a huge amount of data, so it creates fingerprints digesting the original chunks. The software uploads these to the database and compares it with the fingerprints of other documents, checking the originality of the work.Plagiarized documents have been made using the Internet. The text of these documents is copied, so if the system works well, it will show the result of 100% plagiarism. First of these documents is copied from Wikipedia. After the application checked the text, the result was 68%. It is a very disappointing result. After checking more documents with several types of products, it is a fact, that most of the software solutions fail to find all type of plagiarism