Automatic Detection of CSA Media by Multi-modal Feature Fusion for Law Enforcement Support

Schulze, Christian; Henter, Dominik; Borth, Damian; Dengel, Andreas

doi:10.1145/2578726.2578772

Cited by 15 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In this context, Schulze et al [92] proposed a multimodal approach for automatic CSAM detection using visual and audio features. For this purpose, they employed lowlevel features such as skin features, the mid-level sentiment feature SentiBank for images, colour correlograms, visual words as well as audio words for videos.…”

Section: Strategies Based On Motion Audio and Multimodal Analysismentioning

confidence: 99%

Learning Strategies for Sensitive Content Detection

et al. 2023

View full text Add to dashboard Cite

Currently, the volume of sensitive content on the Internet, such as pornography and child pornography, and the amount of time that people spend online (especially children) have led to an increase in the distribution of such content (e.g., images of children being sexually abused, real-time videos of such abuse, grooming activities, etc.). It is therefore essential to have effective IT tools that automate the detection and blocking of this type of material, as manual filtering of huge volumes of data is practically impossible. The goal of this study is to carry out a comprehensive review of different learning strategies for the detection of sensitive content available in the literature, from the most conventional techniques to the most cutting-edge deep learning algorithms, highlighting the strengths and weaknesses of each, as well as the datasets used. The performance and scalability of the different strategies proposed in this work depend on the heterogeneity of the dataset, the feature extraction techniques (hashes, visual, audio, etc.) and the learning algorithms. Finally, new lines of research in sensitive-content detection are presented.

show abstract

Section: Strategies Based On Motion Audio and Multimodal Analysismentioning

confidence: 99%

Learning Strategies for Sensitive Content Detection

et al. 2023

View full text Add to dashboard Cite

show abstract

“…It has also been demonstrated that this approach is able to outperform state-of-the-art porn detection baselines [5] on real-world pornographic and child sexual abuse (CSA) content. Additionally, the compilation of detected ANPs allows to explain detection results to lawenforcement, which in this domain is an important system requirement.…”

Section: Adjective Noun Pairs For Visual Sentimentmentioning

confidence: 99%

Visual Learning of Semantic Concepts in Social Multimedia

Borth

2014

Künstl Intell

Self Cite

View full text Add to dashboard Cite

“…[9], [10]), so far, they have only provided moderate detection rates, due to utilizing single feature descriptions. Also, these tools rely on skin detection techniques, which have been shown to be only marginally discriminative for CSA content detection [11]. Moreover, downloading all files being shared in order to apply such media analysis techniques is clearly infeasible in a P2P scenario.…”

Section: Introductionmentioning

confidence: 99%

iCOP: Automatically Identifying New Child Abuse Media in P2P Networks

Peersman

Schulze

Rashid

et al. 2014

2014 IEEE Security and Privacy Workshops

Self Cite

View full text Add to dashboard Cite

Abstract-The increasing levels of child sex abuse (CSA) media being shared in peer-to-peer (P2P) networks pose a significant challenge for law enforcement agencies. Although a number of P2P monitoring tools to detect offender activity in such networks exist, they typically rely on hash value databases of known CSA media. Such an approach cannot detect new or previously unknown media being shared. Conversely, identifying such new/previously unknown media is a priority for law enforcement -they can be indicators of recent or on-going child abuse. Furthermore, originators of such media can be handson abusers and their apprehension can safeguard children from further abuse. The sheer volume of activity on P2P networks, however, makes manual detection virtually infeasible. In this paper, we present a novel approach that combines sophisticated filename and media analysis techniques to automatically flag new/previously unseen CSA media to investigators. The approach has been implemented into the iCOP toolkit. Our evaluation on real case data shows high degrees of accuracy while hands-on trials with law enforcement officers highlight iCOP's usability and its complementarity to existing investigative workflows.

show abstract

Automatic Detection of CSA Media by Multi-modal Feature Fusion for Law Enforcement Support

Cited by 15 publications

References 16 publications

Learning Strategies for Sensitive Content Detection

Learning Strategies for Sensitive Content Detection

Visual Learning of Semantic Concepts in Social Multimedia

iCOP: Automatically Identifying New Child Abuse Media in P2P Networks

Contact Info

Product

Resources

About