5 ms·

> "Reducing the real-world occurrences for irrational fears doesn't make those fears go away." "You're saying it yourself, these people aren't motivated by stat

> "Reducing the real-world occurrences for irrational fears doesn't make those fears go away." "You're saying it yourself, these people aren't motivated by statistics about abuse, they're frightened of the idea of abuse" We could say the same thing the other way - people up in arms are not frightened by statistics of abuse of a surveillance system, but frightened of the idea of a company or government abusing it. This thread is full of people misrepresenting how it works, claims of slippery slopes straight to tyranny, there's a comparison to IBM and the Holocaust, and it's based on no real data and not even the understanding from simply reading the press release. This thread is not full of statistics and data about existing content filtering and surveillance systems and how often they are actually being abused. For example Skype has intercepted your comms since Microsoft bought it and routed all traffic through them, Chrome and FireFox and Edge do smartscreen blocking of malware websites - what are the stats on those systems being abused to block politically inconvenient memes or similar? Nothing Apple could do would in any way reassure these people because the fears are not based on information. For example your comment: > "We know the risks of outing minors to parents if they're on an LGBTQ+ spectrum." Minors will see the prompt "if you do this, your parents will find out" and can choose not to and the parents don't find out. There's an example of the message in the Apple announcement[1]. This comment from you is reacting to a fear of something disconnected from the facts of what's been announced where that fear is guarded against as part of the design. You could say that the hash database is from a 3rd-party so that it's not Apple acting unilateraly, but that's not taken as reassurance because the government could abuse it. OK guard against that with Apple reviewing the alerts before doing anything with them, that's not reassuring because Apple reviews are incompetent (where do you hear of groups that are both incompetent and capable of implementing worldscale surveillance systems? conspiracy theories, mostly). People say it scans all photos and when they learn that it scans only photos about to be uploaded to iCloud their opinion doesn't seem to change, because it's not reasoned based on facts, perhaps? People say it will be used by abusive partners who will set their partner to be a minor to watch their chats. People explain that you can't change an adult AppleID to a minor one just like that, demonstrating the argument was fear based not fact based. People say it is a new ability for Apple to install spyware in future, but it's obviously not - Apple have been able to "install spyware in future" since they introduced auto-installing iOS updates many years ago. People say it's a slippery slope - companies have changed direction, regulations can change, no change in opinion; nobody has any data or facts about how often systems do slide down slippery slopes, or get dragged back up them. People saying it could be used by bad-actors at Apple to track their Ex's. From the design, it couldn't. But why facts when there's fearmongering to be done? The open letter itself has multiple inaccurate descriptions of how the thing works by the second paragraph to present it as maximally-scary. > "We would also in general like to see more stats about how serious the problem of CSAM actually is" We know[2] that over 12 million reports of child abuse material to NMEC were related to FaceBook messenger and NMEC alone gets over 18 million tips in a year. Does that change your opinion either way? Maybe we could find out more after this system goes live - how many alerts Apple receives and how many they send on. A less panicky "Open Letter to Apple" might encourage them to make that data public, how many times it triggered in a quarter, and ask Apple to commit to removing it if it's not proving effective. And ask Apple to state what they intend to do if asked to make the system detect more things in future. > "their fear is not based on real risk analysis or statistics or practical tradeoffs" Look what would have to happen for this system to ruin your life in the way people here are scaremongering about: - You would have to sync to iCloud, such that this system scans your photos. That's optional. - Someone would have to get a malicious hash into the whole system and a photo matching it onto your device. That's nontrivial to say the least. - Enough of those pictures to trigger the alarm. - The Apple reviewers would have to not notice the false alarm photo of a distorted normal thing. - The NMEC and authorities would have to not dismiss the photo. It's not impossible, but it's in the realms of the XKCD "rubber hose cryptography" comic. Sir Cliff Richard, his house was raided by the police, the media alerted, his name dragged through the mud, then the crown prosecution service decided there was nothing to prosecute. The police apologised. He sued the police and they settled out of court. The BBC apologised. He sued them and won. The crown prosecution service reviewed their decision and reaffirmed that there was nothing to prosecute. His name is tarnished, forever people will be suspicious that he paid someone off or otherwise pulled strings to get away with something; a name-damaging flase alarm which is something what many people fear happening in this thread. Did anyone need to use a generative-adversarial network to create a clashing perceptual hash uploaded into a global analysis platform to trigger a false alarm convincing enough to pass two or three human reviews? No, two men decided they'd try to extort money and made a false rape allegation. People aren't interested in how it works, why it works the way it does, whether it will be an effective crime fighting tool (and how that's decided) or whether it will realistically become a tyrannical system, people aren't interested in whether Apple's size and influence could be an independent oversight on the photoDNA and NCMEC databases to push back against any attempts of them being misused to track other-political topics, people are jumping straight to "horrible governments will be able to disappear critics" and ignoring that horrible governments already do that and have many much easier ways of doing that. > "So in the real world we know that the majority of child abuse comes from people that children already know." Those 12 million reports of child abuse material related to FaceBook messenger; does it make any difference if they involved people the child knew? If so, what difference do you think that makes? And Apple's system is to block the spread of abuse material, not (directly) to reduce abuse itself - which seems an important distinction that you're glossing over in your position "it won't reduce abuse so it shouldn't be built" when the builders are not claiming it will reduce abuse. > "Nobody in any political sphere has ever responded to "think of the children" with "we already thought of them enough."" Are the EFF not in the political sphere? Are the groups quoted in the letter not? Here[3] is a UK government vote from 2014 on communication interception, where it was introduced with "interception, which provides the legal power to acquire the content of a communication, are crucial to fighting crime, protecting children". 31 MPs voted against it. Here[4] is a UK government vote from 2016 on mass retention of UK citizen internet traffic, many MPs voted against it. It's not the case that "think of the children" leads to political universal agreement of any system, as you're stating. Which could be an example of you taking your position by fear instead of fact. > "It doesn't matter what solutions we come up with or what the rate of CSAM drops to, those people are still going to be scared of the idea of privacy itself." The UK government statement linked earlier[2] disagrees when it says "On 8 October 2019, the Council of the EU adopted its conclusions on combating child sexual abuse, stating: “The Council urges the industry to ensure lawful access for law enforcement and other competent authorities to digital evidence, including when encrypted or hosted on IT servers located abroad, without prohibiting or weakening encryption and in full respect of privacy". The people whose views you claim to describe explicitly say the opposite of how you're presenting them. Which, I predict, you're going to dismiss with something that amounts to "I won't change my opinion when presented with this fact", yes? There are real things to criticise about this system, the chilling effect of surveillance, the chance of slippery slope progression, the nature of proprietary systems, the chance of mistakes and bugs in code or human interception, the blurred line between "things you own" and "things you own which are closely tied to the manufacturer's storage and messaging systems" - but most of the criticisms made in this thread are silly. [1] on the right, here: https://www.apple.com/v/child-safety/a/images/child-safety__evd7tla79kqe_large.jpg https://www.apple.com/v/child-safety/a/images/child-safety__... [2] https://www.gov.uk/government/publications/international-statement-end-to-end-encryption-and-public-safety/international-statement-end-to-end-encryption-and-public-safety-accessible-version https://www.gov.uk/government/publications/international-sta... [3] https://www.theyworkforyou.com/debates/?id=2014-07-15a.704.0#g755.3 https://www.theyworkforyou.com/debates/?id=2014-07-15a.704.0... [4] https://www.theyworkforyou.com/debates/?id=2016-06-07a.1142.2#g1155.2 https://www.theyworkforyou.com/debates/?id=2016-06-07a.1142....

danShumway 5y ago
> This thread is not full of statistics and data about existing content filtering and surveillance systems and how often they are actually being abused. It is filled with explanations about why the systems you mention are tangibly different from what Apple is proposing. There is a huge difference between scanning content on-device and scanning content in a cloud. That doesn't mean that scanning content in the cloud can't be dangerous, but it is still tangibly different. There is also a huge difference between a user-inspectable list of malware being blocked in a website and an opaque list of content matches that users can not inspect or debate. There is also a huge difference between a user-inspectable static list and an AI system with questionable accuracy guarantees. And there is a huge difference between a user-controlled malware list that is blocked locally without informing anyone, and a required content list that sends notifications to other people/governments/companies when it is bypassed. That being said, if you want to look at stats about how accurate AI filters are for explicit material in the examples you mention, there are a ton of stats online about that, and they're mostly all quite bad. > nobody has any data or facts about how often systems do slide down slippery slopes, or get dragged back up them There's a lot to unpack in this one sentence, and it would take more time than I'm willing to give, but are you really implying that government surveillance doesn't count as a real slippery slope because sometimes activists reverse the trend? > Minors will see the prompt "if you do this, your parents will find out" and can choose not to and the parents don't find out. There's an example of the message in the Apple announcement[1] You misunderstand the concern. The risk is not the child themselves clicking through to the photo (although it would be easy for them to accidentally do so), it's the risk of that data being leaked from other phones because a friend thoughtlessly clicks through the prompt. > The open letter itself has multiple inaccurate descriptions of how the thing works by the second paragraph to present it as maximally-scary. Where? Here's the second paragraph: > Apple's proposed technology works by continuously monitoring photos saved or shared on the user's iPhone, iPad, or Mac. One system detects if a certain number of objectionable photos is detected in iCloud storage and alerts the authorities. Another notifies a child's parents if iMessage is used to send or receive photos that a machine learning algorithm considers to contain nudity. The only thing I can think of is the word "continuously" which is not strictly inaccurate but could be misinterpreted as saying that the scanning will be constantly running on the same set of photos, and the subtle implication that this scanning will happen to photos "saved", which might be misinterpreted as implying that this scan will happen to photos that aren't uploaded to iCloud. But given that the second sentence immediately clarifies that this is referring to photos uploaded to iCloud, it seems like a bit of a stretch to me to call this misinformation. > people are jumping straight to "horrible governments will be able to disappear critics" and ignoring that horrible governments already do that and have many much easier ways of doing that. Hang on a sec. A little while ago you were calling my fears theoretical, now you're admitting that governments routinely abuse this kind of power. You really think it's unreasonable to be cautious about giving them more of this power? > Does that change your opinion either way? It does not change my opinion, but at least it's real data, so more of that in these debates please. I'm not denying or rejecting the numbers that the UK lists. > A less panicky "Open Letter to Apple" might encourage them to make that data public Holy crud, I would hope this is the bare minimum. Are we really having a debate over whether or not Apple will make that data public? I thought that was just assumed that they would. If that's up in the air right now, then we've sunk really low in the overall conversation about public accountability and human rights. > which seems an important distinction that you're glossing over in your position "it won't reduce abuse so it shouldn't be built" when the builders are not claiming it will reduce abuse. I realize this is branching out on in a different direction, but it sure as heck better reduce abuse or its not worth building. CSAM is disgusting, but the primary reason to target it is to reduce abuse. If reducing CSAM doesn't reduce abuse, it's not worth doing and we should focus our efforts elsewhere. I know this is something that might sound abhorrent to people, but we are having this debate because we care about children. We have to center the debate on the reduction of the creation of CSAM, the reduction of child abuse, and the reduction of gateways into child abuse. Reducing child abuse is the point. We absolutely should demand evidence that these measures reduce child abuse, because reducing child abuse and reducing the creation of CSAM is a really stinking important thing to do. Which leads back to your other note: > does it make any difference if they involved people the child knew? If so, what difference do you think that makes? Yes, it makes a massive difference, because knowing more about where child abusers are coming from and how they interact with their victims makes it easier to target them and will make our efforts more effective. We should care about this stuff. > It's not the case that "think of the children" leads to political universal agreement of any system, as you're stating. I think you misunderstand. Nobody who's willing to bring out "think of the children" as a debate killer has ever dropped the argument because they got a concession. That there are some entities (like the EFF) who are willing to reject the argument as a debate killer and look at it through a risk analysis lens does not mean the unquestioned argument of "one child is too many" is any less toxic to real substantive political debate. > without prohibiting or weakening encryption and in full respect of privacy I don't want to bash on the EU too hard here, but it has this habit of just kind of tacking onto the end of its laws "but make sure no unintended bad things happen" and then acting like that solves all of their problems. It doesn't mean anything when they add these clauses. This is the same EU that argued for copyright filters and then put on the end of their laws, "but also this shouldn't harm free expression or fair use." It means very little to me that the EU says they care about privacy. What real, tangible measures did they include to make sure that in practice encryption would not be weakened? Look, I can do the same thing. Apple should not implement this system, but they should also reduce CSAM. See, I just proved I care about both, exactly as convincingly as the EU! So you know I'm serious about CSAM now, I said that Apple should reduce it. But I predict that you'll "dismiss with something that amounts to 'I won't change my opinion when presented with this fact', yes?" > There are real things to criticise about this system, the chilling effect of surveillance, the chance of slippery slope progression, the nature of proprietary systems, the chance of mistakes and bugs in code or human interception, the blurred line between "things you own" and "things you own which are closely tied to the manufacturer's storage and messaging systems" Wait, hold on. Forget literally everything that we were talking about above. This is, like, 90% of what people are criticizing! What else are people criticizing? These are really big concerns! You got to the end of your comment, then suddenly listed out 6 extremely good reasons to oppose this system, and then finished by saying, "but other than those, what's the problem?"

danShumway 5y ago

> This thread is not full of statistics and data about existing content filtering and surveillance systems and how often they are actually being abused. It is filled with explanations about why the systems you mention are tangibly different from what Apple is proposing. There is a huge difference between scanning content on-device and scanning content in a cloud. That doesn't mean that scanning content in the cloud can't be dangerous, but it is still tangibly different. There is also a huge difference between a user-inspectable list of malware being blocked in a website and an opaque list of content matches that users can not inspect or debate. There is also a huge difference between a user-inspectable static list and an AI system with questionable accuracy guarantees. And there is a huge difference between a user-controlled malware list that is blocked locally without informing anyone, and a required content list that sends notifications to other people/governments/companies when it is bypassed. That being said, if you want to look at stats about how accurate AI filters are for explicit material in the examples you mention, there are a ton of stats online about that, and they're mostly all quite bad. > nobody has any data or facts about how often systems do slide down slippery slopes, or get dragged back up them There's a lot to unpack in this one sentence, and it would take more time than I'm willing to give, but are you really implying that government surveillance doesn't count as a real slippery slope because sometimes activists reverse the trend? > Minors will see the prompt "if you do this, your parents will find out" and can choose not to and the parents don't find out. There's an example of the message in the Apple announcement[1] You misunderstand the concern. The risk is not the child themselves clicking through to the photo (although it would be easy for them to accidentally do so), it's the risk of that data being leaked from other phones because a friend thoughtlessly clicks through the prompt. > The open letter itself has multiple inaccurate descriptions of how the thing works by the second paragraph to present it as maximally-scary. Where? Here's the second paragraph: > Apple's proposed technology works by continuously monitoring photos saved or shared on the user's iPhone, iPad, or Mac. One system detects if a certain number of objectionable photos is detected in iCloud storage and alerts the authorities. Another notifies a child's parents if iMessage is used to send or receive photos that a machine learning algorithm considers to contain nudity. The only thing I can think of is the word "continuously" which is not strictly inaccurate but could be misinterpreted as saying that the scanning will be constantly running on the same set of photos, and the subtle implication that this scanning will happen to photos "saved", which might be misinterpreted as implying that this scan will happen to photos that aren't uploaded to iCloud. But given that the second sentence immediately clarifies that this is referring to photos uploaded to iCloud, it seems like a bit of a stretch to me to call this misinformation. > people are jumping straight to "horrible governments will be able to disappear critics" and ignoring that horrible governments already do that and have many much easier ways of doing that. Hang on a sec. A little while ago you were calling my fears theoretical, now you're admitting that governments routinely abuse this kind of power. You really think it's unreasonable to be cautious about giving them more of this power? > Does that change your opinion either way? It does not change my opinion, but at least it's real data, so more of that in these debates please. I'm not denying or rejecting the numbers that the UK lists. > A less panicky "Open Letter to Apple" might encourage them to make that data public Holy crud, I would hope this is the bare minimum. Are we really having a debate over whether or not Apple will make that data public? I thought that was just assumed that they would. If that's up in the air right now, then we've sunk really low in the overall conversation about public accountability and human rights. > which seems an important distinction that you're glossing over in your position "it won't reduce abuse so it shouldn't be built" when the builders are not claiming it will reduce abuse. I realize this is branching out on in a different direction, but it sure as heck better reduce abuse or its not worth building. CSAM is disgusting, but the primary reason to target it is to reduce abuse. If reducing CSAM doesn't reduce abuse, it's not worth doing and we should focus our efforts elsewhere. I know this is something that might sound abhorrent to people, but we are having this debate because we care about children. We have to center the debate on the reduction of the creation of CSAM, the reduction of child abuse, and the reduction of gateways into child abuse. Reducing child abuse is the point. We absolutely should demand evidence that these measures reduce child abuse, because reducing child abuse and reducing the creation of CSAM is a really stinking important thing to do. Which leads back to your other note: > does it make any difference if they involved people the child knew? If so, what difference do you think that makes? Yes, it makes a massive difference, because knowing more about where child abusers are coming from and how they interact with their victims makes it easier to target them and will make our efforts more effective. We should care about this stuff. > It's not the case that "think of the children" leads to political universal agreement of any system, as you're stating. I think you misunderstand. Nobody who's willing to bring out "think of the children" as a debate killer has ever dropped the argument because they got a concession. That there are some entities (like the EFF) who are willing to reject the argument as a debate killer and look at it through a risk analysis lens does not mean the unquestioned argument of "one child is too many" is any less toxic to real substantive political debate. > without prohibiting or weakening encryption and in full respect of privacy I don't want to bash on the EU too hard here, but it has this habit of just kind of tacking onto the end of its laws "but make sure no unintended bad things happen" and then acting like that solves all of their problems. It doesn't mean anything when they add these clauses. This is the same EU that argued for copyright filters and then put on the end of their laws, "but also this shouldn't harm free expression or fair use." It means very little to me that the EU says they care about privacy. What real, tangible measures did they include to make sure that in practice encryption would not be weakened? Look, I can do the same thing. Apple should not implement this system, but they should also reduce CSAM. See, I just proved I care about both, exactly as convincingly as the EU! So you know I'm serious about CSAM now, I said that Apple should reduce it. But I predict that you'll "dismiss with something that amounts to 'I won't change my opinion when presented with this fact', yes?" > There are real things to criticise about this system, the chilling effect of surveillance, the chance of slippery slope progression, the nature of proprietary systems, the chance of mistakes and bugs in code or human interception, the blurred line between "things you own" and "things you own which are closely tied to the manufacturer's storage and messaging systems" Wait, hold on. Forget literally everything that we were talking about above. This is, like, 90% of what people are criticizing! What else are people criticizing? These are really big concerns! You got to the end of your comment, then suddenly listed out 6 extremely good reasons to oppose this system, and then finished by saying, "but other than those, what's the problem?"