New Proposal Questions License Plate Surveillance Tactics
A proposal known as the Scarecrow system suggests that individuals can bypass automatic license plate recognition (ALPR) systems using unauthorized photography without breaking the law.
This study primarily criticizes the data collection methods employed by ALPR companies, such as Flock, which operates numerous cameras throughout the United States.
As researcher Max Harari commented, Flock “captures and indexes every plate that passes by.” Currently, this system is monitored by around 100,000 surveillance cameras situated in various locations including neighborhoods and parking lots.
Harari emphasized that these cameras collect all plate data, catalog it, and store it in a searchable database—often without a warrant, notification, or public scrutiny.
The Scarecrow project employs a technique that Harari refers to as “adversarial frame pattern optimization.” This strategy generates a grayscale pattern around the license plate, effectively “suppressing detection” while still making it legible to the human eye.
This method reportedly enhances the effectiveness of ALPR evasion, leading to what is termed “total avoidance.” The proposal outlines various distortion tactics including rotation, brightness adjustments, and motion blur, which are said to be effective against most ALPR cameras positioned between 8 to 12 feet away.
To implement this, specialized license plate covers would require photographic evidence of the owner’s plate and would feature unique patterns designed to escape detection by AI systems. Utilizing a 3D printer for these covers might mean that they come with varying depths and specific designs.
Harari expressed a concern regarding systems capable of tracking individuals without transparency or accountability, deeming such practices fundamentally unethical. He mentioned that while he hasn’t directly tested the technology against Flock cameras, preliminary research indicates that it should perform well across different hardware and detection models.
