Are scamera images inadmissable?
Bad security undermines speed camera evidence
Speeding convictions in Australia's state of New South Wales are under attack, following the possibility that the scamera images underpinning speeding cases could have been tampered with. The state's cameras make some AUS$1m a week in fines -- that's about £420,000.
According to one report (see link below), a group of maths enthusiasts has proven that the encryption algorithm, MD5, used to encrypt the images could be broken. This led to a case being thrown out of court by a local magistrate when the state's authorities could not prove that the speed camera's image was definitely taken when the police claimed it was.
No expert could be found to testify that the algorithm made the pictures secure, so the numbers showing time and place of the image could have been changed.
Local motoring group the NMA said that it was crucial the public had confidence in convictions, and called for "a full audit and a review of the system to ensure that it is working appropriately."
The result is that Australia's Roads and Traffic Authority admitted it that could not prove the authenticity of its pictures and that every fine underpinned by pictures taken by speed cameras could be invalid.
The case, seen by Sydney magistrate Lawrence Lawson, has been adjourned for eight weeks so that the RTA can unearth an expert willing to testify that the images are tamper-proof.
It would be interesting to see how much security surrounds images in the UK...
For those of you who might like to know, they use this aglorithm to provide a unique key based on the binary map of a file or piece of text (it doesn't matter) but the claims are that you can get this file, modify it under the covers and put it back and still generate the same key. Very difficult to do right now, and you would have to have got past their network security to do it. Alternatively find the guy who wrote the application and make it an inside job and you're off scott free.
Of course that would be waaaay too costly and time consuming and would make it too inconvenient to keep the cash flowing.
Every time some boiffin invents some new encryption etc, it's only a matter of time before someone cracks it and they have upgrade/re-write the software to close the hole. Just look at online shopping, we're now up to 128bit + encryption, why? Because someone, somewhere has found a hole in everything up to that.
When it's peoples livelihoods on the line, there should be no risk of digital jiggery-pokery!
I'll get down of me now
I'm not sure what the tolerance on speed limits is in the UK, but over here they have been talking about reducing the tolerance to within five percent of the limit before a fine is issued....too bad design standards only require a speedo to be within 10% of the actual speed (and yes, a lot of modern cars are only just within this limit when new).
The UK is obsessed about the moral high ground re the erroreous "speed kills" message and thus the revenue from cameras, but the Australians take pole position at the moment (except the Northern Terriory).
>> Edited by james_j on Thursday 18th August 12:48
annodomini2 said:Ummm - not so sure about this...
Some big holes in the article though, you can't use MD5 to encrypt images for recovery, its a one way algorithm, they could encrypt the images, but once its encrypted they would be unable to recover the image.
>> Edited by annodomini2 on Thursday 18th August 13:56
How about EMC Centera?
A fingerprint is a small data value that is supposed to uniquely identify a piece of data.
The idea is that the same data when fed into a particular hashing algorithm will always generate the same output fingerprint value. If you change just one binary bit of the data however then the hashing algorithm will return a completely different fingerprint value.
The MD5 algorithm has been criticised because it is possible to find two or more blocks of input data that produce the same output fingerprint. This is known as a collision.
However, a computer has to search long and hard for an alternate piece of data that produces the same fingerprint as the original data, and you cannot control or predict what the alternate piece of data will be.
You could replace a scamera image with some random looking garbage that happens to produce the same fingerprint.
However, it is mathematically infeasible that you could replace one scamera image with another legitimate looking scamera image.
I hate scameras, but this case should not have been thrown out of court because of MD5.
puggit said:
annodomini2 said:
Some big holes in the article though, you can't use MD5 to encrypt images for recovery, its a one way algorithm, they could encrypt the images, but once its encrypted they would be unable to recover the image.
>> Edited by annodomini2 on Thursday 18th August 13:56
Ummm - not so sure about this...
How about EMC Centera?
The product you talk about would knock this argument for six and the poor motorist would be paying. However that is a storage product and these people were doing all this up in the application and using it to provide a unique key of the details about when the photo was taken and therefore holding that as a guarantee of authenticity, when if they had only held it as an intrinsic part of the metadata of the image, then it would have been game over and a fine would have been inevitable.
The problem with this is that I have spouted on about something that wouldn't let anyone off with this sort of defense. So I am not really helping the cause of a motorist who uses his or her judgement of the road conditions, time of day, reading the surroundings, etc. to make swift and safe passage to their chosen destination, while at the same time enjoying the thrill of driving.
The key thing here is that the technology used to catch speeding drivers, can be blown away if you take time to understand the technology used.
brigland said:
The key thing here is that the technology used to catch speeding drivers, can be blown away if you take time to understand the technology used.
ourfavouritewebsitemotoringlawwebsite said:
The key thing here is that the laws used to catch speeding drivers, can be blown away if you take time to understand the statute used.
ourfavouritesafetycampaigner said:
The key thing here is that the criteria used to catch speeding drivers, can be blown away if you take time to understand the statistics used.
mylocalscammeratimanager said:
The key thing here is that the sytems used to catch speeding drivers, can be blown away if you do not take time to register your vehicle.
cps said:
The key thing here is that the method used to prosecute speeding drivers, can be blown away if you take time to turn up at court
horridhothatchboy said:
The key thing here is that the method used to catch speeding drivers, can be blown away if you take time to nick a good plate.
captain gatso said:
The key thing here is that the technology used to catch speeding drivers, can be blown away
dvla said:
what vehicle
localjudge said:
The key thing here is that the technology used to catch speeding drivers, is all within an acceptable zone of probability. Conviction stands
My hopes are pie in the sky and then someone eat them
mikesum said:The point is that the existance of 'collisions' introduces an element of doubt regarding the integrity of the evidence. There may be other controls in place that redress that. There are many "evidentially secure" systems that rely on MD5 (and, BTW, there are other hashing algorithms that suffer from collision vulnerabilities) - so these may be challenged in court. An Expert Witness should be able to show that any concerns are groundless - Streaky
MD5 is indeed a one-way algorithm known as a 'hashing' algorithm. It is not used to encrypt data, but rather to create a 'fingerprint' of the data.
A fingerprint is a small data value that is supposed to uniquely identify a piece of data.
The idea is that the same data when fed into a particular hashing algorithm will always generate the same output fingerprint value. If you change just one binary bit of the data however then the hashing algorithm will return a completely different fingerprint value.
The MD5 algorithm has been criticised because it is possible to find two or more blocks of input data that produce the same output fingerprint. This is known as a collision.
However, a computer has to search long and hard for an alternate piece of data that produces the same fingerprint as the original data, and you cannot control or predict what the alternate piece of data will be.
You could replace a scamera image with some random looking garbage that happens to produce the same fingerprint.
However, it is mathematically infeasible that you could replace one scamera image with another legitimate looking scamera image.
I hate scameras, but this case should not have been thrown out of court because of MD5.
joephandango said:
Every time some boiffin invents some new encryption etc, it's only a matter of time before someone cracks it and they have upgrade/re-write the software to close the hole. Just look at online shopping, we're now up to 128bit + encryption, why? Because someone, somewhere has found a hole in everything up to that.
It's not that anything has been 'cracked'. We're now using 128 up to 1024 (and beyond) bit encryption because computers got faster.
Most of the decent secure algorithms are completely open - the code is known publicly both for the public and private areas of the encryption. There's nothing to crack and, without gaining access to both the public and private aspects fo the key you need to use brute force to get through it.
There are no holes in 8 bit encryption - it just doesn't resist guesswork/brute-force attacks for long!
Gassing Station | Speed, Plod & the Law | Top of Page | What's New | My Stuff