Australian police pilot cloud-based AI engine to automate evidence analysis

When investigating a crime, police often end up collecting terrabytes of digital evidence including emails, texts, social media posts, photographs and CCTV footage. Finding connections and patterns in the data is a manual and time-consuming task.

George Nott
Australian_police_pilot_cloud-based_AI_engine_to_automate_evidence_analysis.jpg

When investigating a crime, police often end up collecting terrabytes of digital evidence including emails, texts, social media posts, photographs and CCTV footage. Finding connections and patterns in the data is a manual and time-consuming task.

To help them with this challenge, the Western Australia Police Force (WAPF) is piloting a cloud-based solution developed by Microsoft and partner Modis to automate the analysis of their evidence.

Early results have been promising, police say. When the platform – called Söze, after the antagonist of The Usual Suspects – was applied to an investigation that had been underway for six weeks, it trawled through all the available evidence and identified 18 points of interest to present to the police.

“Some of those things they already knew, which was good. There were a few things that we found that weren’t actually important – but there were other things that we found that were important,” said Anthony Doig, director of innovation for Modis.

The time taken to search digital evidence has also been reduced, Doig says; for example officers are able to find addresses or bank accounts or firearms in hundreds of thousands of images in a few hours, compared with labouring manually over the course of a few months.

It can also quickly find links in the evidence.

“It has found a photo of a person, and then completely different data, say a text message or the mention of that same location the day before, so you can start connecting up those pieces of information and presenting them back to police,” Doig explains.

The platform, based on Microsoft Azure, ingests copies of evidence, the originals of which remain with police. This means “there is no risk that the chain of evidence is compromised” Microsoft says.

Azure Cognitive Services provide the basis for additional platform features, such as translation, facial recognition and text and image analytics.

In the pilot, Doig says WAPF measured productivity improvements in its ability to use digital information in an investigation of 90 per cent.

Detective Inspector Tim Thomas, in charge of covert online operations, digital evidence operations and cyber-crime investigations for the WAPF, called the platform a “paradigm changer”.

“It’s a bit like going from a spoon to a knife and fork. It doesn’t sound like much, but you know, six dinners later you’re going, ‘Wow, this is really making a difference’,” he said.

Profound policing tool

Thomas and his team are currently working with Modis and Microsoft to develop more features.

“We’ll be able to do things like: ‘Show me all of the photographs in police possession taken with this camera’ or ‘show me all the information we have which originated at this place’. No police agency can currently do these things,” he said.

“Well, how is that relevant? Here is a simple example, if we catch a person who has been creating child pornography we will be able to find every picture in our possession which they took quickly and easily. This will allow us to rapidly gain massive insight into where this person has been, who their associates are and what they’ve been doing. Maybe we’ll find victims we didn’t know about before. That’s what I mean about it the effects being profound, and that’s just a simple example,” Thomas added.

The platform will also provide another layer of protection from one of the police force’s biggest fears: overlooking crucial evidence it already has. While that problem used to be due to poor information management, it now comes as a result of simply too much data, Thomas said.

He points to the example of the UK’s ‘Yorkshire Ripper’ case. The West Yorkshire Police were found to have badly mishandled the case; interviewing and releasing the murderer Peter Sutcliffe nine times before he was finally sent to court.

“Put simply, they were unaware of information which they had. Whilst information technology alleviated this problem for a while that grace period is now over. The same technology which saved us is now swamping us with incomprehensible quantities of information,” Thomas said.

“Something has to change, and that’s the point of the platform proof of concept, to help us find the important information we don’t realise we have,” he added.

Australian police pilot cloud-based AI engine to automate evidence analysis

When investigating a crime, police often end up collecting terrabytes of digital evidence including emails, texts, social media posts, photographs and CCTV footage. Finding connections and patterns in the data is a manual and time-consuming task.

George Nott Jun 11th 2019
Australian_police_pilot_cloud-based_AI_engine_to_automate_evidence_analysis.jpg

When investigating a crime, police often end up collecting terrabytes of digital evidence including emails, texts, social media posts, photographs and CCTV footage. Finding connections and patterns in the data is a manual and time-consuming task.

To help them with this challenge, the Western Australia Police Force (WAPF) is piloting a cloud-based solution developed by Microsoft and partner Modis to automate the analysis of their evidence.

Early results have been promising, police say. When the platform – called Söze, after the antagonist of The Usual Suspects – was applied to an investigation that had been underway for six weeks, it trawled through all the available evidence and identified 18 points of interest to present to the police.

“Some of those things they already knew, which was good. There were a few things that we found that weren’t actually important – but there were other things that we found that were important,” said Anthony Doig, director of innovation for Modis.

The time taken to search digital evidence has also been reduced, Doig says; for example officers are able to find addresses or bank accounts or firearms in hundreds of thousands of images in a few hours, compared with labouring manually over the course of a few months.

It can also quickly find links in the evidence.

“It has found a photo of a person, and then completely different data, say a text message or the mention of that same location the day before, so you can start connecting up those pieces of information and presenting them back to police,” Doig explains.

The platform, based on Microsoft Azure, ingests copies of evidence, the originals of which remain with police. This means “there is no risk that the chain of evidence is compromised” Microsoft says.

Azure Cognitive Services provide the basis for additional platform features, such as translation, facial recognition and text and image analytics.

In the pilot, Doig says WAPF measured productivity improvements in its ability to use digital information in an investigation of 90 per cent.

Detective Inspector Tim Thomas, in charge of covert online operations, digital evidence operations and cyber-crime investigations for the WAPF, called the platform a “paradigm changer”.

“It’s a bit like going from a spoon to a knife and fork. It doesn’t sound like much, but you know, six dinners later you’re going, ‘Wow, this is really making a difference’,” he said.

Profound policing tool

Thomas and his team are currently working with Modis and Microsoft to develop more features.

“We’ll be able to do things like: ‘Show me all of the photographs in police possession taken with this camera’ or ‘show me all the information we have which originated at this place’. No police agency can currently do these things,” he said.

“Well, how is that relevant? Here is a simple example, if we catch a person who has been creating child pornography we will be able to find every picture in our possession which they took quickly and easily. This will allow us to rapidly gain massive insight into where this person has been, who their associates are and what they’ve been doing. Maybe we’ll find victims we didn’t know about before. That’s what I mean about it the effects being profound, and that’s just a simple example,” Thomas added.

The platform will also provide another layer of protection from one of the police force’s biggest fears: overlooking crucial evidence it already has. While that problem used to be due to poor information management, it now comes as a result of simply too much data, Thomas said.

He points to the example of the UK’s ‘Yorkshire Ripper’ case. The West Yorkshire Police were found to have badly mishandled the case; interviewing and releasing the murderer Peter Sutcliffe nine times before he was finally sent to court.

“Put simply, they were unaware of information which they had. Whilst information technology alleviated this problem for a while that grace period is now over. The same technology which saved us is now swamping us with incomprehensible quantities of information,” Thomas said.

“Something has to change, and that’s the point of the platform proof of concept, to help us find the important information we don’t realise we have,” he added.