CSAM Detection via Age and Sexuality Component Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning models struggle to detect child sexual abuse material (CSAM) in streaming media in real-time due to the illegal nature of CSAM, making it impossible to use standard supervised training approaches, and existing methods often detect CSAM after it has already been displayed, failing to prevent its appearance.
Innovation Solution
A machine learning model comprising an age component, a sexuality component, and a combination component is trained to detect CSAM in streaming media by analyzing faces for age and sexuality, generating outcomes that indicate the presence of CSAM, with a prediction component predicting CSAM in future frames to prevent its display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If standard supervised training approaches are used to train machine learning models for CSAM detection, then the model can be trained with labeled data, but it is impossible to obtain and use CSAM data for training due to its illegal nature
Solution Approach 1:
The model is segmented into multiple independent components: an age determination model that analyzes facial features to estimate age, and a sexuality detection model that identifies sexual content. These components are trained separately on legal, non-CSAM data, avoiding the need to train on illegal CSAM materials while still enabling accurate CSAM detection through combination of their outputs
Solution Approach 2:
The age determination model acts as an intermediary component that provides age probability information to the CSAM detection logic. This intermediary approach allows the system to infer CSAM presence indirectly through age assessment and sexuality detection rather than directly analyzing CSAM content
2Loss of time
If existing CSAM detection methods are used, then CSAM can be detected, but detection occurs after CSAM has already been displayed, failing to prevent its appearance
Solution Approach 1:
The system performs preliminary detection on individual frames using the age and sexuality models before assembling them into a complete video. This allows early identification of potential CSAM content, enabling preventive actions such as blocking or blurring to be taken before the content is fully displayed to the user
3Productivity
If real-time CSAM detection is implemented in streaming media, then immediate detection and prevention is possible, but the system must process and analyze video data continuously without interruption
Solution Approach 1:
The video stream is segmented into individual frames that are processed independently and in parallel by the age determination model and sexuality detection model. This frame-level segmentation enables real-time processing throughput while maintaining system modularity and reducing computational complexity compared to analyzing entire video sequences
Data Source
AI summary
There is provided a method of training a machine learning model, comprising: extracting faces depicted in image(s), creating an age training dataset comprising records, each including a face and a ground truth label indicating whether the face is below a legal age, training an age component on the age training dataset for generating a first outcome indicative of a target face from the target streaming being below the legal age, creating a sexuality training dataset comprising records, each including image(s) and ground truth label indicative of sexuality depicted therein, training a sexuality component on the sexuality training dataset for generating a second outcome indicative of sexuality depicted in the target streaming, defining a combination component that receives an input of a combination of the first outcome and the second outcome, and generates a third outcome indicative of child sexual abusive materials (CSAM) depicted in the target streaming.