The Technical Way Chat Control Scans Your Messages

i keep seeing people argue about chat control who clearly don't understand how it actually works technically. so let me explain the scanning mechanism in detail, from someone who's actually looked at the technical specifications and the academic papers behind them.

this is going to get technical. if you just want the "what should i do" version, skip to the end.

client-side scanning: the core mechanism

chat control uses a technique called client-side scanning (CSS). here's how it works, step by step:

1. a detection database is pushed to your device. this is a database of "hashes" — mathematical fingerprints of known illegal content. the database is maintained by organizations like NCMEC (national center for missing & exploited children) and is regularly updated.

2. when you send a message containing an image or file, the scanning software computes a hash of that content on your device. this happens before encryption.

3. the hash is compared against the detection database. if there's a match, a "match report" is generated and sent to the platform, which forwards it to law enforcement.

4. if there's no match, the message proceeds normally. in theory, nobody sees the content except you and the recipient.

the key technical detail: the scanning happens on your hardware, using your battery, consuming your bandwidth for database updates.

the hashing algorithms

two main approaches are used:

perceptual hashing (PDQ, PhotoDNA)

these algorithms generate a "fingerprint" of an image that's resistant to resizing, cropping, color changes, and compression. even if you modify an image significantly, the hash should still match.

PDQ (Perceptual hash for Detection Quality) is meta's open-source implementation. PhotoDNA is microsoft's proprietary version. both are used in the EU's scanning infrastructure.

the problem: perceptual hashes have a false positive rate. studies have shown rates between 1 in 10,000 and 1 in 100,000 depending on the threshold. sounds low? the EU sends billions of messages daily. even a 1 in 100,000 false positive rate means thousands of innocent people flagged every day.

neural hash (apple's abandoned approach)

apple developed a more sophisticated approach using neural networks to generate hashes. it was more accurate but also more computationally expensive. apple abandoned it after massive backlash.

the EU's technical working group has been evaluating neural hash approaches for chat control 2.0.

the detection database problem

here's something most people don't realize: the detection database is a black box.

• you can't inspect it
• you can't verify what's in it
• you can't audit it for false positives
• the organizations that maintain it operate with minimal transparency

the database could theoretically contain fingerprints of political content, copyrighted material, or anything else — and your device would scan for it without your knowledge.

the evasion problem

the technical community has identified numerous ways to defeat client-side scanning:

• minor image modifications can defeat perceptual hashing
• steganography — hiding data within other data — bypasses content-based scanning
• custom encryption layers applied before the app's encryption
• platform switching — using platforms that don't implement scanning

but here's the dark side: every time evasion techniques are published, the detection algorithms get updated. it's an arms race, and the defender (the scanning system) has a fundamental advantage because it can update silently through automatic database pushes.

what this means practically

the technical reality is:

1. client-side scanning will catch some genuinely illegal content
2. it will generate false positives affecting innocent people
3. it will be expanded beyond its original scope (this is historically guaranteed)
4. it can be evaded by sophisticated actors, meaning it primarily affects ordinary people
5. it cannot be independently audited or verified

i wrote a much more detailed technical analysis with diagrams and code examples: Technical Guide to Chat Control Scanning

understanding the technical mechanism is the first step to understanding why so many security researchers and cryptographers oppose this system. it's not about protecting criminals — it's about building surveillance infrastructure that will be abused, because surveillance infrastructure is always eventually abused.