2 ms·
What part of comparing checksums of images on your device to KNOWN child pornography checksums is "undermining your interests"? This was not some Google-style
by theshrike79 4y ago
What part of comparing checksums of images on your device to KNOWN child pornography checksums is "undermining your interests"?
This was not some Google-style hotdog or not algorithm that was using "AI" to guess if a photo had a nude person on it or not.
- saagarjha 4y ago1. What's to guarantee that the checksums are of child pornography? 2. What's going on with the "fuzzy comparison"?
- theshrike79 4y agoThe database is generated by the NCMEC[0]. I trust them to have a rigorous process for adding stuff to it. [0] https://en.wikipedia.org/wiki/National_Center_for_Missing_%26_Exploited_Children https://en.wikipedia.org/wiki/National_Center_for_Missing_%2...
- saagarjha 4y agoA lot of the concern is that others may not trust the NCMEC, or that they don't trust that other images won't be added in whatever ends up on your device.
- mindslight 4y agoFirst, the huge possibility of false positives. They're not "checksums", but rather "perceptual hashes", which are nowhere near as solid as a cryptographic hash. When you multiply a small chance of a false positive across hundreds of millions of deployed devices, you end up with many false positives - false positives that result in people being characterized as "child abusers". Second, the overall aim of the technology also targets legitimate or inadvertent activities, that the law has also unjustly criminalized - say taking family pictures of your kids, 17 year olds sexting, or someone trolling you. Once again unjustly branding people as "child abusers", this time with "evidence" to back up the narrative. But even in the case of a proper match to the NCMEC list for its bona fide purpose, that still is undermining the phone's owner's interests. That's the philosophical contour, even if you personally wish to brush it aside with the desire to catch people looking at evil images. Our society respects similar longstanding privacy boundaries, even if it does end up helping some "bad people". It's really not in the NCMEC's interest to respect any of this, as their foundational dynamic is that there is a horrible thing happening in the world, and they must do everything possible to stop it. That kind of advocacy is certainly needed, but we shouldn't just accept their desires as if they're an unbiased neutral party.
- theshrike79 4y agoYes, the internet went on a "generate false perceptual hash positives" binge. Everyone just forgot that in the Apple system you don't get automatically banned and reported to the authorities if you get a match (or five). The matches get manually checked by an actual human, who will see in 0.2 seconds that it's not actual CP, but a highly distorted picture of a cat or some gibberish. Zero action will be taken. Also: the scanning was only done if iCloud is enabled. If iCloud is enabled, the EXACT SAME scan could be done in the cloud. Why are people not objecting to this with the same fervour?