Fighting Child Exploitation with AI and Data Science – AFP and Monash University

Fighting Child Exploitation with AI and Data Science – AFP and Monash University

October 27, 2019 0 By Stanley Isaacs


– Ten years ago we were dealing
with a couple of phones, maybe some CDs. And now like everything else
electronic, with the internet, with hard drives we’re
looking at thousands of times more data. But there aren’t thousands
of times more police. For a person to be charged for
possessing child exploitation or radicalization material for
example, a human investigator has to review it and say, “yep, I say that is child exploitation”. It has an emotional impact. Each person shows it
differently but it is there and sadly, I guess the biggest problem is often you don’t realise the impact until it’s already happened. – One of the issues in dealing
with image classification is that you need to train
your classification algorithms on data, on real data. In this case this real data is sensitive, it’s illegal to possess in many cases and it’s distressing to view. So we’ve come up with a
system called a data airlock in conjunction with Data61
which enables researchers to test and train their
algorithms on this data without actually being
exposed to it themselves. We think this is a world first initiative and it’s a key contribution of AiLECS. – What’s unique about this research is the collaboration between
law enforcement and university. There are maybe two or three
other places in the world where you’ll get access to
the data, the motivation and the resources to do this stuff. The results have been strong enough that it’s kicked off Project
Stonefish in the federal police which is an automation
piece to enable the training of classifiers out at Monash
as part of the AiLECS lab without requiring any more work by digital forensics members,
investigators or police. – In general, this
technology has application beyond child exploitation, in particular counter-terrorism investigations
or any investigation where highly distressing and
offensive imagery or video is likely to be encountered. – The AiLECS lab is a really good example of where we’re using AI to
have a positive social impact on society. And we realised from that that when we looked
across Monash University we had hundreds of similar
examples of using AI in a wide variety of fields but
all with a mission statement of AI for social good. And so we decided to
bring all of that together under one banner and we called it the Monash
Data Futures Institute.