That's the same argument that police departments use "Well we'll only use it to capture child abductors and stolen cars and stuff".
But the truth is that there just aren't that many real child abductors out there (especially with known license plates -- and when they are known, it tends to be a child-custody dispute rather than a stranger-abduction), and meanwhile a huge database of every driver's whereabouts is being amassed with no good controls over who owns the data or what it's used for.
I'm surprised companies aren't offering the service for free in return for the database. There must be huge marketing possibilities... John Doe drives past this oil change place twice a day, let's send him a coupon, or better... John Doe's car goes to Jane Smith's house twice a week and parks there for two hours, let's send his wife an ad for a private investigator.
I'm getting into home automation at the moment, so a use case that I think I deleted from the article was just letting you know that your mum or someone had pulled up in the driveway. License plates vs. a known list of visitors. Or as is happening here in some of our suburbs, "Alert: A group of 4 men have arrived in a stolen V8 Commodore"
It's only creepy if it's done on a targeted, individual basis, which it won't be.
As it is - in many countries you legally have no expectation of privacy in a public place, so I have no problem with en-masse ANPR. If the police eventually develops a very dense sensor network that can track every car's movement across every intersection then there would be too much information recorded for me to personally worry about it - and I still believe it would help with crime in general. The usual reactionary bullet-points based around the usual bogey-men of terrorists and child-abduction are weak and I think distract from more meaningful and impactful, but sedate arguments: e.g. it would reduce the need for inherently-dangerous hot pursuits to follow a vehicle as the cameras would do that anyway. Uninsured vehicles could be identified immediately before they get stopped for a broken taillight or involved in a collision. And cynically: APR systems could be used to compute a car's average-speed between points to determine if the car broke the speed limit).
I don't feel mass ANPR is comparable to other bulk-surveillance schemes like Internet snooping, because I believe that we do have an expectation of privacy regarding what goes through our home connections, and acting on Internet surveillance is inherently subjective - whereas ANPR can indicate if a car is known to be stolen or uninsured, that's not something that's open to interpretation.
IIRC recently there was a city in CA that mailed everyone who drove down a certain street a letter in an obvious colored envelope saying "Dear, Whoever, consider yourself informed that prostitutes go here, kthanksbye". This is a public road that people drive down to get from A to B.
The cost comes from compliance, audits, paperwork. Trials, PoCs, etc.
You also need to prove security and a long term lifecycle for the solution. Those open source packages - are they being updated? Who validates patches? etc etc.
Those are moats, once you're able to deliver the above, you're set for a long time.
Good luck!
A mistake like this can and has destroyed lives.
http://thisiscriminal.com/episode-18-695bgk/
> Police officer John Edwards was patrolling a quiet neighborhood in Bellaire, Texas when he saw an SUV driven by two young African-American men. It was just before 2am on December 31, 2008. Edwards followed the SUV and ran the license plate number. His computer indicated that the SUV was stolen, and Edwards drew his gun and told the two men to get down on the ground. It wasn’t until later that he realized he’d typed the wrong license plate number into his computer. He was off by one digit. By the time he realized his mistake, one of the men had already been shot in the chest at close range.
I had hoped I'd made it clear enough, but by no means am I recommending something I cobbled together in an hour on the couch be rolled out as a competitor to BlueNet. I mean shit, I'm web scraping their database and using compressed footage from cheap dashcams. The algorithm is also untrained.
I was, on the other hand, hoping to promote healthy discussion and pose some good questions related to IT procurement here in Australia. There needs to be a happy medium between what I've wired up and an $86M solution.
For a full disclaimer, many many years ago I did work on various Victoria Police IT projects in a previous company. While our projects were delivered on time and on budget, we did hear some horror stories about what the multinational consulting firms were charging.
Obviously complicated requirements exist but I think there's a way of embracing this with some skilled workers
I think that the digital service is trying this a bit
> It found public servants were too afraid to make major changes to IT procurement and were not talking with other departments to avoid duplication.
> "A fear of external scrutiny of decisions — such as through Senate estimates and audits — leads to a low-risk appetite and a culture where it is 'not OK to fail'," the report said.
> "This means that old and familiar ICT solutions are preferred to newer and more innovative, but perceivably riskier, solutions."
Nobody in the APS gets fired for hiring IBM, and nobody in IBM gets fired for screwing up major government projects. http://www.abc.net.au/news/2016-10-25/turning-router-off-and...
Don't ask me why, but something tells me there was a bit more to the actual Victoria Police project than just the image-scanning component. But the title makes for good clickbait, at least.
Not that I'm saying $86M is justified - rather that I can start to see how things might add up.
Well, if we're just checking to see if a car is registered stolen, that's hopefully a very small database. A simple hash table should be sufficient for a yes/no decision. But God only knows what goes on inside the DMV, and getting that data out into a rapidly queryable state could be tricky.
Google reckons there's only 4.5 million registered vehicles in Victoria (.au) - with 6 character [A-Z0-9] = 36bit/5byte plates that's only gonna need ~22 meg to store them all and you've got a few bits for "flags" - even if 10% of them are listed as stolen at any time, you'd only need a couple of meg for the "stolen car list" - even without handwaving away the overhead for the data structure, this seems a reasonable solution. If you want the make/model/colour info you could then just query the remote db when you get a hit for a plate with the "stolen" flag set (or perhaps the "vehicle of interest" flag).
Of course nobody's gonna build it that way, because then you lose the data set of plate locations you can own without needing to ask for consent. This will, of course, store GPS location and timestamp of every plate is acquires, and if they're being particularly audacious, the actual imagery captured as well. That's how you make this a $80mil project... :-/
(Of course private industry already has this - tow trucks carry alpr rigs and sell the data to repo agencies in the US, I'm told...)
I can understand that the $86 million seems like a lot until you realize the kinds of capabilities that an automotive camera system like has to comply with and all of the integration needed with other software/systems/services and the fact that this cost is likely a total cost over many years as it sounds much more sensational. Just the initial report from Deloitte cost the government $115,000.
In their first trial in 2014-2015, which lasted 15 months, with only 6 cars outfitted, ended up identifying and impounding 240 stolen cars and increased revenue from additional tickets by hundreds of thousands of dollars per month. That's pretty impressive! Now they're rolling out to not just 6 cars but to 220 total, which in theory should improve policing efforts.
References: http://www.caradvice.com.au/350313/victoria-police-to-consid... http://www.heraldsun.com.au/news/law-order/victoria-polices-...
I honestly believe it is examples like this that clearly show how quickly the world is changing, and how massive organizations are a thing of the past.
You don't need thousands of employees and $86M, you need a couple of smart people and a weekend.
Uber and Air BnB have to some degree done this already, I have no doubt the exact same thing will happen in health, education, transportation and many, many other aspects of our lives.
"Why did this system cost 86 million dollars" would be much more informative. It could be that the whole thing is a waste, but that would require an actual analysis.
Seriously? You jumped the shark from interesting tech, to think of the children it's ok to go all Orwell here.
First App I'd make. Cheating spouse app. Is your partner or Ex hiding from you. Pay here for tracking info. Part of which goes to the people collating data.
To the author: please spend some time and try to analyze the problem first. I don't think you even get the problem right, or know what open source is. A free api IS NOT an example of OPEN SOURCE.
I'm using an open source license plate recognition system (OPENALPR), taking the produced registration number, and THEN querying the (yes, closed source) records database via web scraping. All tools used to achieve this, bar where the records are kept, are open source (openalpr, cheerio, node-horseman etc).
At nearly 400,000 AUD per vehicle, the question should be what better opportunities to help people did the Australian Govt pass up in order to engage this project?
At a quick glance, this thing is going to need power, probably from the car battery. You don't want it to catch fire and kill the occupants, so there's going to be R&D in power supplies, material etc, along with regulation and testing.
Just the other day the trigger happy cops shot a couple of people at a swingers party.
Despite the tragic result, the officers were unanimously acquitted of any criminal wrongdoing.
https://www.usatoday.com/story/news/nation/2014/05/24/suprem...
"Mistake? Ha ha, we don't make mistakes."
http://www.independent.co.uk/news/world/americas/policeman-b...
In the end I understood, when I came to understand, as other in this thread have pointed out, that there's support, rollout, training, documentation, project management, planning and it goes on and on. The technical solution can be just a small company of providing a sustainable solution to an organization.
Having said that, you should be packing this up and selling it.
Let's assume that your system is wrong 1 out 10 times or 100 times. And they're are 1 million scan a year. that's going to result in 100,000 people pulled over wrong. For a cost of let's say 860K(assume your outfitting the system, training people, data storage etc.).
Let's say your going to need 4 to 5 sigma better performance to get that. Does that mean 10x more effort? 100x more effort? Think data collection, storage, possibly state of the art algorithm...
what if the data gets hacked.
All I got out of your article is there's already no way to tell the difference in cost between an toy system and enterprise, government system.
BTW, it was still a cool toy project...
A related story for this discussion: QUT students design a $500 cloud-based census server four times better than IBM’s $9 million system [1]
[1] http://www.news.com.au/technology/online/qut-students-design...
I have two objections, which the article doesn't seem to fully address.
First, do we know that the server the students designed actually met the spec handed to IBM? Often times a lot of the complexity of a project comes from the interaction of a few features. It's quite possible that the version the students provided didn't actually do the hard stuff. We know from the article that they specifically didn't address any security/privacy ramifications of sitting running in the cloud.
We also don't know how many similar projects were attempted and failed. Sometimes things just come together, way quicker and cheaper than should have been expected.
Many federal projects are now under a single sign on portal that works very well. Our taxes take less than an hour to fill out as most details are pre-filled.
And speaking for NSW, the RTA/RMS/Whatever-they're-called-these-days (our equivalent to the DMV for US readers) is almost fully functional without an in-person visit.
As to the device it needs to do more than just read licence plates. It must compare them with something which means it needs to get over the air updates for a list of licence plates. So, now you need some back end system to get important lists of licence plates from various other systems. As to the device, how much coverage does it get aka does it need multiple cameras.
And on it goes.
He laughed and said that there is so much electronics inside that the battery would run down during lunch.
I didn't think to ask why they didn't turn off the electronics. I also didn't ask what they do at the end of the day.
I helped the ACLU stuff the envelopes. I consider that some of my best work.
All of History says otherwise. All surveillance mechanisms have been used to individually track dissidents, personal enemies, ex-spouses, etc. What makes this different?
If the police eventually develops a very dense sensor network that can track every car's movement across every intersection then there would be too much information recorded for me to personally worry about it
The availability of large databases (ex. DNA) have been shown to increase the number of innocent people caught in the net, since the probability of false positives increases proportionally and cops start relying on running wide searches over regular investigation techniques.
it would reduce the need for inherently-dangerous hot pursuits to follow a vehicle as the cameras would do that anyway
Only if you have a complete panopticon, with no gaps, which in a wide enough are will be difficult, if not impossible.
It wouldn't really reduce the need for the infamous hot-pursuits because often in these hot pursuits, the car is stolen so the driver is not who the car is licensed to.
With a large enough ANPR network, there is no need to chase the car - so the driver won't necessarily panic or act irrationally to evade the police - leading to a higher probability of recovering the vehicle in good condition, and a successful non-dramatic apprehension of the suspects as it makes it easier to catch them by surprise, for example.
I hope I never live in a society where the driver of a stolen car can expect to be apprehended no matter where he stops because he'll be tracked by a camera network so ubiquitous that he can't escape it.
Because such a ubiquitous network means that all of my travels are being tracked and recorded too - which will likely include facial recognition, so there really will be no escaping Big Brother... and I don't trust any government with so much power and information over citizens to be benign.
- who has access and are they using it for official business?
- what if it's hacked?
- who is it shared with and who do they share it with after that? private companies for analysis?
I should have said rarely.
Every LEO in the country is taught to shoot only when deadly force is necessary and to shoot for the center of mass (and continue shooting until the suspect is down).
Shooting with intent only to inflict an immobilizing but not-life-threatening wound is inherently wrong and has been held actionable by the courts. If the LEO had the luxury and time to even consider such a tactic, then aim and fire, then by definition no innocent life was in imminent danger, and therefore the use of deadly force was excessive.
And if you take another look at your source, you'll see that there is no stated intent to only wound the suspect in that German case.
A good study of such situations is the Massad Ayoob book In the Gravest Extreme.
If you come at cops with a knife I think it's "reasonable" they shoot you. What is not reasonable in the case of US cops is that they seem to shoot even non dangerous people and get away with it.