Inference of a compact representation of sensor fingerprint for source camera identification

作者：

Highlights：

• An effective framework for de-noising and compressing sensor pattern noise (a form of device fingerprint) is proposed.

• A compact and discriminative representation can be obtained directly from a high-dimensional SPN.

• A novel training set construction method is designed to minimize the impact of various interfering artifacts in training samples.

• The proposed methods can be used as a post-processing framework applied after any SPN extraction methods for effective source camera identification.

摘要

•An effective framework for de-noising and compressing sensor pattern noise (a form of device fingerprint) is proposed.•A compact and discriminative representation can be obtained directly from a high-dimensional SPN.•A novel training set construction method is designed to minimize the impact of various interfering artifacts in training samples.•The proposed methods can be used as a post-processing framework applied after any SPN extraction methods for effective source camera identification.

论文关键词：Image forensics,Source camera identification (SCI),Sensor pattern noise (SPN),PCA de-noising

论文评审过程：Received 30 November 2016, Revised 28 August 2017, Accepted 18 September 2017, Available online 27 September 2017, Version of Record 10 October 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.09.027