Erwin van Eijk explained how justice first tries to build a hypothesis. The prosecution states that the suspect actually committed a crime. The hypothesis from the defense is that he didn't. The NFI's task is to evaluate the evidence against the hypotheses and get the most probable answer out of this.
The NFI does this independently by trying to fulfill the task as best as science can help. The NFI uses all the tools that science is applying to get their answers. In the end, astronomy is also an hypothesis being falsified or supported.
Camera recognition has faults during production and those faults can be seen later on in the imagery taken, whether it is video or pictures. Together with the eScience Center in Amsterdam, the NFI is working on a project to automate image recognition. Basically, you are looking for noise in the image and you want to see if other pictures have the same noise distribution.
With that you can help either party to find or disprove that a picture was taken with a certain camera or was taken with the same camera. The last option is for instance necessary when you have a picture of a suspect with his wife and the attorney claims that this picture was taken with the same camera as all the pictures containing the crime, while the defense is claiming that it wasn't.
This is a very computationally intensive task. The NFI is using HPC to crack that. The same goes for cracking passwords or cracking encryption in general, which is also something being done at the NFI facilities.
Most people want to know how much resources the NFI has for a task but unfortunately, Erwin van Eijk cannot tell this.
The other tasks performed by the NFI are more involved with Big Data than with HPC. The average case is very big. The average case amounts to four terabytes of material consisting of seized hard drives, seized cameras or seized phones. Those need to be processed within the deadline set by the law. This is a different deadline than the one for an article you want to publish inNature. If you miss the article deadline, it will only hurt your own ego. Missing the deadline for court basically has legal consequences.
Therefore, the NFI invests quite some engineering time to process the data. A smaller amount of time is possible using all the techniques known in the HPC market but also in the Big Data market. Most of those problems are I/O bound instead of compute bound. This has its own problem space compared to the very large compute tasks.
In general you can say that detectives don't know what they are looking for. They are looking for something that is not allowed, but where it is and how they can find it, is very unknown upfront. It is even worse: you don't know if a certain piece of evidence has no value whatsoever to the case before you looked at it. You have to analyze the entire piece of evidence, you have to evaluate the entire hard drive in order to be able to say that there is no value in it and it can go back to the suspect or whoever it was seized from.
This is not funny when it is a four terabyte hard drive because just copying a four terabyte hard drive is going to take something like twelve to fifteen hours, depending on the brand, model and speed. But still, the NFI's goal is to perform the computing as fast as the original medium can bring the data to the processing facilities.
In that way, the NFI knows that there is no way the answer could not be presented any sooner that at that point of time.
For that reason, the NFI is also doing other projects with university or research institutes in the Netherlands. The NFI is not very big. Most of the NFI staff consists of forensic investigators or forensic scientists. In the end, computer science is just a means of getting to answer the hypotheses. Not all of the NFI people are working on HPC or large processing problems.
It is very convenient that the Netherlands has quite excellent research facilities and research institutes you can cooperate with and talk with to exchange ideas. They can also work on ideas the NFI has because they think it's an interesting problem. This has direct relevance to society in general. If you can reject a piece of evidence sooner, it also means you can let someone go. The suspect goes to an investigative judge but you can let him go.
This is something people tend to forget. If you can get the results sooner, it also means that someone who is not involved in the case, can go.