Forensic identification of resampling operators: A semi non-intrusive approach

详细信息	查看全文 | 推荐本文 |

• 作者：Gang Cao^a ; ^b ; ¹ ; ^{06112056@bjtu.edu.cn} ; ^{gcaocn@gmail.com} ; Yao Zhao^a ; ^b ; ^{yzhao@bjtu.edu.cn} ; Rongrong Ni^a ; ^b ; ^{rrni@bjtu.edu.cn}
• 关键词：Digital forensics ; Semi non-intrusive ; Software forensics ; Resampling
• 刊名：Forensic Science International
• 出版年：2012
• 期刊代码：104_03790738
• 类别：med
• 出版时间：10 March, 2012
• 卷：216
• 期：1-3
• 页码：29-36
• 文件大小：713 K

摘要

Recently, several new resampling operators have been proposed and successfully invalidate the existing resampling detectors. However, the reliability of such anti-forensic techniques is unaware and needs to be investigated. In this paper, we focus on the forensic identification of digital image resampling operators including the traditional type and the anti-forensic type which hides the trace of traditional resampling. Various resampling algorithms involving geometric distortion (GD)-based, dual-path-based and postprocessing-based are investigated. The identification is achieved in the manner of semi non-intrusive, supposing the resampling software could be accessed. Given an input pattern of monotone signal, polarity aberration of GD-based resampled signal's first derivative is analyzed theoretically and measured by effective feature metric. Dual-path-based and postprocessing-based resampling can also be identified by feeding proper test patterns. Experimental results on various parameter settings demonstrate the effectiveness of the proposed approach.