Data Forensics Constructions from Cryptographic Hashing and Coding

Abstract

Data forensics needs techniques that gather digital evidence of data corruption. While techniques like error correcting codes, disjunct matrices and cryptographic hashing are frequently studied and used in practical applications, very few research efforts have been done to rigorously evaluate and combine benefits of these techniques for data forensics purposes. In this paper we formulate unifying algorithm, data and security models that allow to evaluate and prove the security guarantees provided by direct forensic encoding constructions from these techniques and suitable combinations of them.We rigorously clarify the different security guarantees provided by using these techniques (alone or in some standard or novel combinations) for both data at rest and data in transit. Our most novel construction provides a forensic encoding scheme that allows to detect if any errors were introduced by corrupted data senders, does not allow data intruders to detect whether the data was encoded or not, and requires no data expansion in a large-min-entropy data model, as typical in multimedia data.