Artificial Intelligence / Machine Learning
Artificial Intelligence, AI, and machine learning is increasingly being used in child sexual abuse investigations, by helping to recognise, categorise and triage material. It is a rapidly developing technology, which shows a lot of promise for this type of investigative work. Unlike hashing technologies, AI classifiers have the potential to recognise new and previously unclassified child sexual material.
One machine learning technique is artificial neural networks (ANN), which are based on efforts to model information processing in the human brain. The adage is that Artificial Intelligence is learning without instruction or programming.
ANN are adaptive systems that change based on external or internal information that flows through the network, i.e. the system is learning, and the networks infer functions from this learning. This is important in systems where the complexity of the data or task makes the design or function by hand difficult.
ANN learn and adapt through assessing data, and in order to draw the right conclusions they must train on high volumes of quality data. An AI application is only as good as the data on which it has trained. If the data is flawed the system will draw the wrong conclusions and become inefficient, or unhelpful. This is why, for optimum results (output) it is crucial to have high quality data and to structure the training in such a way that the system draws the right conclusions.
AI shows great future potential for finding, analysing and removing child sexual abuse material online.
AI classifiers are being developed at speed to assist with law enforcement investigations, and aside from law enforcement, industry is also developing AI applications to detect child sexual abuse material. One example is Google’s AI classifier that can be used to detect child sexual abuse material in networks, services and on platforms.
There is also a clear case for businesses and organisations to use this technology. When a child sexual abuse image is detected in an IT environment, other files on the device can be searched with the help of an AI classifier. It is also possible to schedule these types of searches in the IT environment.
Strengths and limitations
AI classifiers are unique in that they have the ability to detect previously unknown material. This increases the scope for detection, and identification and safeguarding of previously unknown victims. The technology is developing fast, and will continue to revolutionise the fight against online child sexual abuse exploitation.
However, an AI classifier is not a hundred percent reliable, and it will make mistakes. It is, in essence, only as good as the data that it has been provided with. If the data is flawed or too limited, the AI will draw the wrong conclusions, and even with the most high-quality data, it will still make mistakes. Hence, it still relies heavily on human verification to ensure that the AI classification is right.
Another limitation is that when an AI classifier makes mistakes, it is usually very difficult, if not impossible, to backtrack why it has made a particular mistake.
Artificial Intelligence / Machine Learning is one of many technologies that can be applied by businesses to stop child sexual abuse material. In the last section of the NetClean Report 2019 we presented an overview of technologies and methods available to businesses to stop child sexual abuse material. The articles were a revision and abridgement of longer and more technically detailed articles, published here. In a series of blog posts we will compare the different technologies and show how they complement each other.