Hacker News Reader

Comment by dev_dull | Hacker News Reader

Are they worth it as a target? With an RPI and openalpr I can easily build my own license plate reader in one evening. This technology is so far out of the bag already.

I’m critical of people like me (millennial, working in tech) and I think I have good reason. I hear so many bizarre technological “solutions” to what are ultimately policy issues. If we spent half that time instead lobbying our representatives we would be in a much better place as a society. Can you name your state rep? How about you write them a little today rather than succumb to cynicism or spitballing tech.

View on HN

Reading the license plate is trivial, its processing, and doing the advanced queries that make ALPR solutions real.

Sure you could create a bunch of pi cams able to handle a few reads per second, but then say you want to know what plates have traveled with your target plate between multiple camera sets to see if someone is being followed, or you have cell phone pings and want to search all cameras in the radius of that tower for a plate.

The reading isn't hard, its fully solved, the economical and real-time searching of plates and evidence compilation is the much harder problem.

It’s true. There is also a lot of data cleaning and grouping to do. Building sighting histories gets computationally intensive with scale.

> Building sighting histories gets computationally intensive with scale.

Can you talk more about why this is difficult?

My very naive assumption is it would be pretty easy to partition/index the data by plate number and location/time and use something like Storm or Spark (?) to run various types of queries.

Sure. I have a SaaS, EasyALPR.com that uses ALPR for parking enforcement that I've been working on for a few years.

I'm near a major release actually, replacing my previous products with something called Parking Enforcer, which I believe is the best mobile app / vehicle grouping tool in market. I focus on business parks with 300-2000 parking spaces to patrol. It has been in beta for about 9 months.

I have been working algorithms related to ALPR data set matching a lot, primarily in Python.

I'm not familiar with storm / spark. However, one issue is that that license plate reads are not 100% accurate. So you are looking for a fuzzy version of the plate.

Its possible for collections of plates to sort of fuzz-out as incorrect members become more distant from previous ones, causing new matches to join groups they should not. You can write stuff to handle this but the original point was ~"this problem is in analysis not ALPR" which I agree with.

As far as computation, a new plate may or may not have a group to join. If it has been seen before, you need to look for the group "most likely" to be the same car. This can mean iterating over a large data set to look for the highest probable group.

There are some tricks to cutting down on the data set under consideration. For example it is much easier to only look at possible sightings from this past week in this area than every sighting from everywhere. (which when assembling a national database may be necessary)

But in my experience, even in the tens of thousands of plates, doing grouping requires task queues and tricks for quick identification and notification of matches. A watchlist (blacklist) is might be short, but the grouping task over months of data can be long.

Some plates are VERY similar but not the same car! Sometimes the ALPR camera forwards photos of fences or vehicle grills that are not license plates at all, and those must be ignored but not at a threshold where good data is thrown out. When bad stuff makes it to the user, they have to be easily disposed of if they make it through.

Data cleaning things can require additional capabilities of analysis, like breaking apart improperly grouped vehicles, yet ensuring they do not rejoin "bad" groups. Lots of details to handle, and to an end user, it is painfully obvious when "matches" are not correct. They expect magic.

One other note is that I found building an efficient and useful model architecture for these purposes to be challenging. There are more details in how the raw data comes in from ALPR scans that have to be handled before you even reach an individual "Sighting."

Question, from a technical perspective, how much of your fuzzy/unclear plate problem do you think will be solved by advances in cameras? In the context of a car with 4 to 6 cameras mounted on it roving a parking lot checking plates. Already the video output from the hdmi port on a GoPro hero7 black is considerably better than what a $4000 camera could do just 3.5 years ago, and it costs $349... Or various setups I have seen with modified 22 megapixel Sony mirrorless cameras used for GIS orthophotos from drone platforms.

Hard to say. FWIW, much of the time the exact plate gets read. But it also matters what has been trained. Some states have weird vanity plates. So it not purely an optics problem.

If I could I would have all my customers scan plates with the iPhone XS. But many work with 7 or even less.

Also introducing a lot more collected data to dedupe is a thing as well. One of the first algos I worked on was just realizing in short term new dats we already had a car and to not try and treat the same car like it was another car.

Are there ALPR companies with a workflow where the truly unrecognizable plates go in a review queue for image recognition by a human offshore somewhere, in a low cost location? I'm thinking of the standard call centre salary for people with an average level of education in second/third-tier cities in India, in Pakistan, in Bangladesh, etc.

It that I’m aware of. But it is certainly possible. There is just a lot of data.

There are some tools that focus on this “fuzzy deduplication” problem:

Senzing https://senzing.com/

Tamr https://www.tamr.com/

Thank you I will look at these.

> I hear so many bizarre technological “solutions” to what are ultimately policy issues.

Tech is empowering, much moreso than playing politics.

Consider something like the Kafkaesque nightmare that is applying for (and keeping) food stamps. It doesn't need to be complicated (nor should it be!), but you can either try to convince elected officials to make the poor a priority and fix the process or roll up your sleeves and write a script to complete the forms in triplicate, generate mailing labels with delivery confirmation, remind users of deadlines and pull phone records to prove the social worker never called like they said they did in the denial letter.

Or you could petition, harass, bribe and cajole your way into enacting change, and have it all overturned with a change in administration.

In some part these technical solutions exist to fix people problems. Look at the internet itself-- where problems exist (a country's politicians/dictator makes the nation unroutable), you don't wait for a coup, you route around it.

But also in some part these solutions are just modern rent-seeking, so...

>you don't wait for a coup, you route around it.

No, we’re still waiting for a coup in China, North Korea, and on and on for the citizens to access the internet.

> Or you could petition, harass, bribe and cajole your way into enacting change, and have it all overturned with a change in administration.

This is the reason why technological solutions are popular among the Silicon Valley crowd. No matter what, political solutions are plagued by human emotion and self-interest, and thus they become sticky, "corrupt", and slow. Technological solutions are subversive of that structure at the least, and a force multiplier in others.

The computer will generally do what you tell it to do. You can spend hours of effort on something and get a deterministic result that will do the thing. You can spend your whole life in politics and get nothing out of it because the entrenched power structures won't let it happen.

The computer will do what people tell it to do, so it’s just as plagued by human emotion and the self-interest of those who program it.

It’s just a different form of the same corrupt politics.

The biggest problem with technology is this delusion that it somehow isn’t a reflection of human failings.

If we at least admitted that then we’d be able to reason about it responsibly.

More technology to process forms just enables the bureaucracy to create more forms, in the same way that computer programs have gotten more fat off of the increased power and capability of modern devices. It makes the process shitty for everybody else who don't want to pay for, can't pay for, or aren't covered by the nifty tech. Exhibit A being TurboTax.

Technology is a band-aid, not a solution, and technology can also be used to make progress towards solving the problem decay, or actively worse.

I think "covering up gangrene" works well with the band-aid analogy

> Consider something like the Kafkaesque nightmare that is applying for (and keeping) food stamps.

I have long thought that this particular process was intentionally difficult as a means of helping to weed out people who aren't really in need.

> Are they worth it as a target? ... This technology is so far out of the bag already.

The point is that if you have access to their systems you can affect what is reported. For example you could add a bogus plate to the data stream, or remove one, or perform a substitution. I think we can all imagine cases were doing such a thing might be useful to someone.

Others have also pointed out that it’s possible that this company has some particularly interesting recognition technology, but I agree with you that this is really a second order issue.

I was thinking that if you can see the logic behind their license plate recognition algorithm you can engineer modifications to license plates that would make them unrecognizable by the police. For example, adding a weird black square on a few of the numbers that look fine to the naked eye but for whatever reason screw their scanners up.

excellent point!

Yep, something like this:

https://www.theverge.com/2019/4/23/18512472/fool-ai-surveill...

Isn't the article stating that the hackers are posting the actual license plate scans of people crossing the border? That seems like a big deal to me.

Politics is about convincing as many people as possible to vote your way, and to make a big difference, you have to do it at scale. Changing people's minds is hard. Much like sales, it also has downsides like making a pest of yourself.

It's important, but I am a bit frustrated by shallow encouragement that makes politics sound like it's easy. It's like saying "get a job" rather than providing actually helpful resources and training to find a job.

I think the reason why this is significant is not so much that the proprietary files were leaked, rather the amount and breadth of personal information exposed along with the insinuation that this company's network (who maintains large amounts of personal data of unwitting participants) is not secure.

I bet a little digging and we will discover that this network was hopelessly undefended and their software is horribly riddled with holes and poor security practices.

It was obviously done to impact the company as an entity rather than to target individuals or the government. It's quite obvious, to me, that the attackers wanted to dump a large liability in someones lap and perhaps for good reason. If this company can't secure the data it collects about innocent civilians then it shouldn't be allowed to collect it.

Perceptics is a pretty big player in the license plate camera market. This breach is extremely embarrassing for them and will doubtless put their future work with government in jeopardy as they become a political target. I wonder if this hack was perpetrated by China as retribution for the Huawei blockade? The Chinese perhaps have an arsenal of such breaches ready for release when the right politician needs a nudge.

>> Can you name your state rep? How about you write them a little today rather than succumb to cynicism or spitballing tech.

Yes I can name (all) of my state's representatives. No I'm not going to write them. Your letter and sending information get filed under an equivalent of "dissenters". That's kind of the entire point here.

What makes you think the object of the attack was to duplicate their technology?

I would be more interested in the security implications. Hackers now have access to the source code; what kind of attack surfaces does this yield against customers who purchased the compromised software?

How does lobbying work? Is it backed by some kind of power other than the eager generosity to provide free policy consultation?

if you have the ML that parses the plates, you could make adversarial plates that it couldn't find

Reply on news.ycombinator.com