Biometrically linking document leakage to the individuals responsible

Document Type

Conference Proceeding

Publisher

Springer Verlag

School

School of Science

RAS ID

29543

Comments

Alruban, A., Clarke, N., Li, F., & Furnell, S. (2018, September). Biometrically Linking Document Leakage to the Individuals Responsible. In International Conference on Trust and Privacy in Digital Business (pp. 135-149). Springer, Cham. Available here

Abstract

Insider threats are a significant security issue. The last decade has witnessed countless instances of data loss and exposure in which data has become publicly available and easily accessible. Losing or disclosing sensitive data or confidential information may cause substantial financial and reputational damage to a company. Whilst more recent research has specifically focused on the insider misuse problem, it has tended to focus on the information itself – either through its protection or approaches to detect leakage. In contrast, this paper presents a proactive approach to the attribution of misuse via information leakage using biometrics and a locality-sensitive hashing scheme. The hash digest of the object (e.g. a document) is mapped with the given biometric information of the person who interacted with it and generates a digital imprint file that represents the correlation between the two parties. The proposed approach does not directly store or preserve any explicit biometric information nor document copy in a repository. It is only the established correlation (imprint) is kept for the purpose of reconstructing the mapped information once an incident occurred. Comprehensive experiments for the proposed approach have shown that it is highly possible to establish this correlation even when the original version has undergone significant file modification. In many scenarios, such as changing the file format r removing parts of the document, including words and sentences, it was possible to extract and reconstruct the correlated biometric information out of a modified document (e.g. 100 words were deleted) with an average success rate of 89.31%.

DOI

10.1007/978-3-319-98385-1_10

Access Rights

subscription content

Share

 
COinS