Secure similar detection for documents

Abstract EN

Data similarity detection is an important for many applications such as document file management, document searching, plagiarism prevention, copyright protection.

Most of the researches does not rely the preservation of privacy in detecting the similarity for documents because it is considered that the content of the document is general, which is reduced the use in certain applications that require the preservation of data when the detect the similarity, e.

g.

, the conference submissions are treated as confidential and not revealing them to other program (in the process of similar document detection).

Over the past few years, cryptologists have created protocols that preserve the privacy and protection of data while detecting similarities but it remains not secure.

In this paper, we evaluate the similarity between two parties (Alice and Bob) without knowing any information about the content of their files, with maintaining efficiency and repairing the security problems of previous works.

The cosine similarity is used to measure the similarity between the vectors of the document.

Our proposal was applied on a real dataset through several experiments conducted to demonstrate the value and efficiency of proposed schemes in practice.