A Statistical Approach for Identifying Videos of Child Pornography at Crime Scenes

Abstract

The identification of child pornographic files at crime scenes can take a pedophile to jail immediately. However, this type of live analysis is a difficult task, since a computer storage device can actually store millions of files. NuDetective Forensic Tool was developed for this purpose and provides satisfactory results in the identification of these files, especially images, through automatic nudity detection. However, this tool does not analyze video file content, which is a temporal media, requiring a more complex analysis. In literature, the main studies of automatic pornography detection in videos do not have the primary purpose of being quick, and are not focused in the detection of child pornography. This work presents a new strategy for automatic detection of child pornographic videos. The new approach uses nudity detection on extracted video frames and statistics to perform the automatic identification of these illegal videos. Experiments showed that our new approach is quick enough to be performed at crime scenes with detection rates around 85%, bringing new benefits for forensic examiners in the identification of child pornographic files.