3 ms·

I don’t like this line of argument: it has weak knees for 3 reasons. 1. They question the “1 in 5 children are abused” figure. That is fair: it seems tremendou

I don’t like this line of argument: it has weak knees for 3 reasons. 1. They question the “1 in 5 children are abused” figure. That is fair: it seems tremendous at first sight. But the real figure of 1 in 10 that they reveal is very close and similarly shocking… 2. They emphasize that children know their abusers. But then imply that scanning chat messages cannot help in this situation. There is an obvious retort: abusers can abuse children in their local community and share images of their abuse online. Catching that sharing can identify them and stop them. (I cannot see any statistics on how many, among abusers, actively share things about it online. Intuitively, it would be a low percentage, but that argument is not used.) 3. They calculate that police surveillance orders were only issued for 0.1% of abuse cases, implying that they do it so little because it is ineffective. But the obvious retort is that they do it so little because they have no grounds to do so. “A chat sent from this person’s phone tripped the detector” is a reasonable justification. (I expect there are some much more obvious ways to have that justification which are poorly tapped, like confidential school reporting. But that is not part of the argument.) The fundamental argument should be that abusers are listening to the EU too. They will gear up. There is nothing anyone can detect in encrypted base64 sent by chat or other means. Meanwhile, any weakness created by this system will be turned into abuse imparted on the innocents, like sharing to an unwitting victim a zipped adversarial image of a prairie whose pixels were flipped until it triggered the detection. Overall, that, along with a possibly inadequate false positive rate given the rarity of real abusers 1) actively sharing and 2) doing so insecurely, it is possible that this system would cause more victims than it would help.

josefx 4y ago
> “A chat sent from this person’s phone tripped the detector” is a reasonable justification. A black box detector that similar to a trained drug dog will bark on command. Last time a secret internet block list pushed by von der Leyen was leaked it became clear that they wont limit this stuff to child protection and will do their best to keep the actual criteria used secret. Just mentioning the name von der Leyen in a chat might get you on a list. It is times like these where I would love to have the GDR back, a place where we could just drop of all the crazy Stasi wannabees.
- espadrine 4y ago
  > A black box detector that similar to a trained drug dog will bark on command. Absolutely. There is never any description given of the required false positive rate, or the way it will be measured and enforced. A false positive rate of 1% across ~450M people, to find 3M abusers (assuming each abuser averages two victims based on the article’s figures) of which only 1% exchange images insecurely, would mean that 0.8% of detections would be valid. The police cannot chase 125 dead ends for a single arrest; they just don’t have that kind of budget. They already give up on finding bike thieves even when the bikes are GPS’ed within town, and that would have a much higher probability of success per investigation.

josefx 4y ago

> “A chat sent from this person’s phone tripped the detector” is a reasonable justification. A black box detector that similar to a trained drug dog will bark on command. Last time a secret internet block list pushed by von der Leyen was leaked it became clear that they wont limit this stuff to child protection and will do their best to keep the actual criteria used secret. Just mentioning the name von der Leyen in a chat might get you on a list. It is times like these where I would love to have the GDR back, a place where we could just drop of all the crazy Stasi wannabees.

espadrine 4y ago
> A black box detector that similar to a trained drug dog will bark on command. Absolutely. There is never any description given of the required false positive rate, or the way it will be measured and enforced. A false positive rate of 1% across ~450M people, to find 3M abusers (assuming each abuser averages two victims based on the article’s figures) of which only 1% exchange images insecurely, would mean that 0.8% of detections would be valid. The police cannot chase 125 dead ends for a single arrest; they just don’t have that kind of budget. They already give up on finding bike thieves even when the bikes are GPS’ed within town, and that would have a much higher probability of success per investigation.