Hackers Got Inside a Flock Camera
Unmasking the Eye: Inside the Flock Camera Breach
By Dhruv Mehrotra & Joseph Cox | Security | September 16, 2026
In a bold move to expose the inner workings of modern surveillance, a hacker collective has physically seized a Flock Safety camera, extracted its internal data, and leaked the findings to WIRED and 404 Media. This breach provides a rare, transparent look at a system that the company has long insisted is impenetrable secure.
“Why just destroy them when we can reverse engineer them and find the secrets of those spying on us?”
— Member of the collective stegan0gram
📊 The Scale of Surveillance
The dumped data reveals a staggering volume of information captured in a very short window. The hackers recovered logs and videos that paint a picture of constant, high-resolution monitoring.
| Metric | Value |
|---|---|
| Timeframe | 21 Days |
| Unique Vehicles | 50,000 |
| Total Images Captured | 1.6 Million |
| Avg. Images per Vehicle |
Using LaTeX to express the capture density:
🛠️ Technical Breakdown: How it Works
The device is essentially a specialized Android-based computer. It utilizes a processor comparable to a mid-range smartphone and runs approximately 20 proprietary Flock applications.
The Data Architecture
The hackers discovered the storage was split into different sections, or partitions. While Flock claimed the data was encrypted, the hackers found a loophole:
vendor: An unencrypted partition.media: An unencrypted partition containing the encryption key.- Encrypted Storage: The actual footage, which was unlocked using the key found in the
mediapartition.
# Conceptual view of the recovered partition structure
/dev/block/bootdevice/by-name/
├── vendor/ # [UNENCRYPTED] System configs
├── media/ # [UNENCRYPTED] Contains key.bin <--- KEY FOUND HERE
└── data_encrypted/ # [ENCRYPTED] Vehicle videos & stills
The Capture Process
When the camera detects motion, it doesn't just take one photo. It triggers a rapid-fire sequence:
- Quantity: Most vehicles trigger ~28 images; some trigger over 100.
- Exposure: The system uses varying exposure settings to ensure both the license plate and the surrounding environment are clear.
- Object Detection: The AI identifies more than just plates. It explicitly flags:
- Vehicles
- License Plates
- Bicycles
- People
- Specific Graphics (e.g., bumper stickers or an American flag patch on a motorcycle bag).
🌐 The "National Network" Controversy
Flock's primary selling point is its interconnectedness. Data isn't just for the local police; it's often shared across a massive web of agencies.
Case Study: Alpharetta, Georgia WIRED discovered that data from this single city was accessible to over 2,000 different organizations, including:
- Local and State Police
- University Campus Security
- Airport Authorities
- The Office of Inspector General for the federal General Services Administration (GSA)
Reported Misuse
The partnership with 404 Media highlighted several disturbing applications of this network:
- ICE Collaborations: Local officers performing searches for Immigration and Customs Enforcement, even in "sanctuary" jurisdictions that forbid such data transfers.
- Medical Policing: A Texas officer utilized the national network to track a woman who had performed a self-administered abortion.
🛡️ Security Claims vs. Reality
The breach contradicts previous assurances from Flock Safety. In early 2025, researcher Jon “GainSec” Gaines identified flaws allowing root-level access.
Flock's Stance:
- Acknowledged the flaws.
- Denied severity (claimed physical access was required).
- Claimed footage is transient and deleted quickly after cloud upload.
The Hacker's Reality:
- Physical access was obtained.
- Root-level access was achieved.
- Encryption keys were recovered from the device.
- Thousands of videos were successfully dumped.
🏁 Conclusion
While some activists have resorted to simply smashing cameras—leading to arrests and the deployment of 3D-printed "decoy" cameras by police—the stegan0gram collective has shifted the strategy toward reverse engineering. By "liberating" the hardware, they have exposed that the system is far more invasive and less secure than the company admits.
