Flock…

Is having a ‘bad’ month…

Hackers Got Inside a Flock Camera. Its Data Shows How the System Really Works

A hacker collective pulled down a Flock camera and dumped its data. The files included thousands of videos and logs showing that the device captured 1.6 million images of 50,000 vehicles in 21 days.

Full article, HERE from Wired and from Slashdot, HERE. h/t Steve

The really scary part of the analysis is this-

The software running on the camera explicitly detects people, something which is typically overlooked in discussions around Flock cameras. When it spots a person, it records where they appear in the image and how confident it is in the detection.

And HERE is an article from 404 Media about what the police are doing to ‘justify’ their searches (or not)…

Ironically, I saw two ‘new’ Flock installations have gone up on I-44 between Wichita Falls and Burkburnett, TX when I went up that way on Sunday for breakfast. I found this really interesting, since Gov Abbott had signed to stop Flock cameras on Texas highways 28 August 2026, with state money, HERE.

Now I’m wondering ‘who’ is paying for those cameras to get put up there???

Oh yeah, and Flock’s ‘vaunted’ security… not so much… sigh…

I REALLY don’t want to live in a police state! One of the reasons I moved all the way out here.  Dammit…


Comments

Flock… — 21 Comments

  1. Hey Old NFO;

    I heard a lot of the flock camera are a cooperative between local jurisdictions and private companies, both split the “cut” as it were. Do I agree with this, HELL NO. I have heard a lot of flock camera’s being placed around gun ranges to they can track who goes shooting…this one really bothers me, sure there is no national gun registry…..”But why Mr. Smith is your car going to the gun range on Willow Ave 2 times a month?”….THe stasi in the government will always try a work around the bill of rights.

  2. They probably found a legal work around where the private company owns it and LE has a subscription.

    And I’m not surprised that their security stinks. It is nearly always an afterthought.

    Next thing, we’ll find that they keep data well beyond 21 days and sell it to lots of companies.
    I’m surprised that hasn’t happened already.

  3. Counter story
    The cameras have no more storage than your Ring doorbell. It’s all sent to a data center.

    Disclaimer: I have not dissected a Flock camera to verify either story. Occam’s Razor: the scary story is a psyop.

    • Addition: they can process the images in the data center for whatever they want to find. Remember when Google never looked at your pictures but “would you like to see all your pictures with that person in them?”

  4. For those opposing the installation of public surveillance tech, another site noted (with photos) that many of these installations violate either state required ownership / installation authority labels, and / or are installed contrary to state road safety laws.
    (Similar to mailbox setback standards)
    Check your State laws, regulations, and road safety standards.
    Locate and identify those state and local authorities who are permitting these intrusions, “just for future reference.”

  5. One place old west values still exists is Moffat County, CO. The county decided to not renew their Flock contract and asked Flock to remove their equipment. Crickets, so the Sheriff had his deputies removed the equipment. Said equipment is safely stored for Flock to retrieve by the end of the year.

  6. Bob- Yep, at least one of those in Florida, because a friend of mine took the pics of the cameras at the entrance to the gun range.

    Jonathan- Yep, ‘that’ is the question- ‘Where’ is the data kept for 21 days???

    Steve- If that is true, why was 21 days of data found on the camera?

    John- Yeah, I still haven’t gotten a response to the FOIA I filed… sigh

    WSF- Yay!!!

  7. Upon some research (Brave AI)
    So yes, officially captures data, does some processing in camera (model dependent). The hacker reveal is single source but verified by others.

    Being the skeptic I asked if the hackers could have false flagged the software part (AI’s can write code really fast these days, so possible) but apparently very unlikely (below).

    tl;dr Bottom line, hacker reveal looks legit. Welcome to 1984.

    Theoretically possible, but practically implausible. Here’s why:

    The selfie test is the killer detail. WIRED ran the extracted models against a new photo of one of their own reporters — an image the hackers had no way of knowing about or pre-fabricating for. The model correctly detected the person. If the hackers had fabricated the models, they’d need to build a working object-detection network that happens to correctly identify a specific journalist’s face in a photo taken after the hack.

    The fabrication burden is enormous. To fake this, the hackers would need to:

    Build a functional CV model matching the Qualcomm MSM8953’s NPU capabilities
    Embed it in a coherent Android 8.1 app structure alongside ~19 other Flock apps
    Generate fake log files with consistent timestamps, bounding boxes, and confidence scores across 21 days of activity
    Make the burst-capture behavior (28 frames/vehicle) match the hardware’s actual frame rate
    Get DDoSecrets to distribute the full filesystem publicly for anyone to audit
    Get two independent editorial teams to both verify it hands-on
    Corroborating context makes it even less likely:

    Flock’s own marketing confirms on-device AI for its PTZ cameras (Guardian Mode, People Detection)
    Flock’s patent explicitly mentions cameras detecting “race”
    404 Media confirmed in December 2025 that Flock uses overseas gig workers to train ML models on people/vehicle classification
    The “Vehicle Fingerprint” feature Flock markets implies on-device scene analysis
    The one legitimate caveat the mlq.ai article raises: the exact camera model, serial, and firmware version weren’t identified, so results may not generalize across all Flock configurations. But that’s a generalizability question, not a fabrication one.

    Bottom line: The combination of the selfie test, the publicly auditable raw filesystem, the hardware-consistent behavior, and Flock’s own prior disclosures makes a fabrication scenario effectively ruled out.

  8. To expand on the surveillance, after Grass Valley, California cancelled the Flock cameras and removed them, Flock then went in and replaced them.
    Link
    So it looks like if Flock cannot get money from the city, they are getting money thru some other way. Is that by recording images and selling them to third parties and if so who are they selling to?

    Also to the point made above about the Gov’t working around the Constitution & laws, that’s where Parallel Construction comes from. I’ve read Apologetics saying that because Flock is a “private company”, the Constitution and laws don’t apply to them. And since the police are getting the information from Flock, it’s being provided by “publically available sources” and therefore doesn’t require a warrant or any other judicial review.

    That leads to the stories of police arresting & jailing people because Flock “misidentified” them.

  9. As I live in a Police State (Australia) I can assure you you don’t want to be living in one and my state (Queensland) is less obnoxious than the rest of the country. The police write the gun laws in Australia and speed limits are rigidly enforced. Two mph over (3 km/hour) gets a ticket.
    Moved to 5 acres 6 miles out of reasonable size town (15000 people) and off grid.

    • “Two mph over”
      Anybody know the margin of error claimed for a speedometer? Mine reads 2 or 3 mph fast according to the many roadside radar-on-a-trailers I’ve driven past. Except that one which keeps insisting I’m going 17 mph over; pretty sure it ain’t right as everyone is passing me.

      • Australian speedometers are allowed to read from zero error to 10% fast.
        There is no excuse on new cars as most have GNSS and a speedometer can be real time calibrated when a good GNSS signal is available so should be accurate all the time.
        GNSS velocity is done straight from the doppler shift, not from difference between two positions. Super accurate and low noise to usually less than one cm/sec.

    • And the speed limits in Oz are very slow for the comparable areas in the rest of the western world. The vehicles are very horsepower / engine size limited, and the distances vast. Basically makes travel out of local areas much harder to accomplish. Sydney to Canberra takes about an hour longer than it should. 4 lane divided highway speeds comparable to 55MPH in the US. Qld resisted the covid lockdown schemes longer than any other Oz state, but still had lots of pressure from Canberra. Annastacia Palaszczuk can choke on a bag of rocks for all the harm she did.

  10. Steve- Thanks!

    Eyrie- I know… Spent time in Perth and Sydney… sigh

    Robert- Yeah, Ohio back in the day. 56 in a 55 going DOWNHILL! Grrrr…

    Ag- Oh, me like!

    Stretch- Before or after, or both???

  11. Flock doesn’t care. They aren’t going anywhere. The criminals in power like them way too much to get rid of them. Thus the need for extralegal vigilante response to them.

  12. Many jurisdictions eliminating Flock cameras are quietly replacing them with Axon cameras.
    Axon is the rebranded Taser company so Axon already has a built in list of customers. These cameras collect a lot more than Flock.
    An interesting read is the patent application for the Flock cameras.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.