4 ms·

> literally security by obscurity You've said that a lot, but there is no obscurity here. Or security. If you actually want to use this for serious spam filter

by kpcyrd 11y ago

> literally security by obscurity You've said that a lot, but there is no obscurity here. Or security. If you actually want to use this for serious spam filtering, you'd discard all emails without a valid + suffix with a whitelist.

logicallee 11y ago
I simply don't think you're writing in earnest, since your proposal is quite nuanced and does address my issue, proving that you exactly understand it. But gmail does not support the filter you propose. ("discard all emails without a valid + suffix with a whitelist.") Also even if you were serious you would know that if I did manage to get gmail to filter all email without a valid + suffix from a whitelist, I would then have a choice: I could give out meaningful suffixes, like me+amazon, me+ebay, me+facebook, or meaningless hihg-entropy suffixes. But if I give out meaningful suffixes, then spammers can just try the email address I give out, replacing the portion after + that identifies them, with likely candidates instead. They are unlikely to succeed, but this is still security by obscurity. If we move to true security, picking high-entropy strings to white-list, that means I now need to manually track the relationship between the strings and the party they identify. Google should track that relationship, not me. It's a single key-value store. That's it. Associate me+ebay with squirehamboard so that I can give ebay "squirehamboard@gmail.com" and have it behave exactly as though I gave ebay "me+ebay@gmail.com" today. I think you're better than this. Be better. I have faith in you. You understand the issue. If you were working for Google you could solve it within 7 minutes (proof of concept). It's a single key-value store. I think you know very well the definition of security by obscurity. Be better than this. Be better!
- newjersey 11y ago
  How would I solve it without parking n number of aliases which are now forever unavailable for anyone else to use? You could do what I do and just create a separate google email address (mine literally has the word spam in it) that feeds into your primary Gmail but goes straight to all mail and is marked as read by default. It is funny when some websites reject my email address because the word spam in the address. Perhaps Google could make this process simpler by allowing autogenerated email addresses that feed into the real primary email without the user having to create a second email address. I think this would benefit Google as the user data isn't scattered around in multiple Google accounts. So for example I can request a new alias and Google would make one like ezJVoWkJFBPvMm98xhOUsi7X52l06RNblLXhhCFse3nwdkpCW7VUkIO7zgJqQDd at gmail Dot com and forward that to my real email address. If I ever respond to any email that came to that address, make it look like my email address is that string above. Now since it is just a random 63 (or longest allowed by Google) char alias that the user didn't choose, it is unlikely that it will be anything someone wants as their email address. Is this along the lines of what you were thinking?
  - logicallee 11y ago
    You think extremely similar to me!!! For the sites that reject mailinator and its aliases, I give them a gmail address that literally has the word spam in it, just as yours does. But I just check it for any registration I need to confirm, I don't forward it to my main account. But it literally has the word spam in it, just like your spam gmail :) Interestingly, I've never once had any site reject that due to having the word spam in it - but maybe because it's my second-tier possibility only for use on sites that reject mailinator. I thought it would be likely that some sites would reject it for that reason, but I just haven't had that issue. What you've written: >Perhaps Google could make this process simpler by allowing autogenerated email addresses that feed into the real primary email without the user having to create a second email address. Is precisely what I suggest. Specifically, all of the current me+comment@gmail.com (which feeds this email address to the me@gmail.com email address, marking it with your comment)should be kept, it's just that instead of giving Bob's Spam Factory the email address "me+bobspam@gmail.com" I first request a high-entropy version of "me+bobspam", and get something similar to your suggestion: >ezJVoWkJFBPvMm98xhOUsi7X52l06RNblLXhhCFse3nwdkpCW7VUkIO7zgJqQDd@gmail.com and give Bob's Spam Factory that address instead. Then Bob's Spam Factory can't just take out the part after the + to get my primary email address. However, you do not need 63 random characters in order not to worry about taking possibilities from someone!!! (Not even close). Given a dictionary of 100,000 words, 3 random words have 100,000 * 100,000 * 100,000 = 1,000,000,000,000,000 or 1 quadrillion possibilities. It's vanishingly unlikely that anyone would ever pick the same strings. This is what I meant by "entropy" which is a technical term. To see this in action, consider that I came up with the "high-entropy" string squirehamboard for my example. I haven't checked yet, but it is extremely likely (99.99%) that I'm the first person to have ever written the words squireharmboard one after the other, and that if I hadn't written it, nobody would write it in the next thousand years (except exhaustively listing all possibilities, but I mean actually picking it or writing it by hand). Let's check. Let's google squirehamboard: https://www.google.com/search?q=squirehamboard https://www.google.com/search?q=squirehamboard - as you can see, nobody in the history of the Internet has written that short phrase except me. It's not the only example I had. I also listed in my original comment buffaloaerypyrite - again, it is extremely likely that I'm the only person to have ever written buffaloaerypyrite in the history of the Internet (I used a high-entropy free association method to come up with these words), and that if I hadn't written it, nobody would in the next thousand years. Let's check: https://www.google.com/search?q=buffaloaerypyrite https://www.google.com/search?q=buffaloaerypyrite Again, I'm the only person to have ever used the string buffaloaerypyrite on the entire Internet in any context. (Of course, some web server could decide to exhaustively list all three-word possibilities, since my personal mental dictionary that I free-associated through to get the terms Buffalo, Aery, Pyrite, as well as Squire, Ham and Board, from does not have 100,000 words in it - I know more like a few thousand words personally. It's possible for someone to list buffaloaerypyrite in a huge (multi-multi-terabyte) collection of passwords. But I don't quite require Google's handles to be as strong as passwords. They should just be long enough not to step on anyone's toes who are making real, normal email addresses. In practice 3 random words from a huge dictionary is more than enough. In practice you will never have any conflict with any real email address anyone chooses. But yeah, you get the basic idea. You just go overkill on the length of password that you think needs to be chosen :) 3 random words given by Google are more than enough...

by kpcyrd 11y ago

logicallee 11y ago

I simply don't think you're writing in earnest, since your proposal is quite nuanced and does address my issue, proving that you exactly understand it. But gmail does not support the filter you propose. ("discard all emails without a valid + suffix with a whitelist.") Also even if you were serious you would know that if I did manage to get gmail to filter all email without a valid + suffix from a whitelist, I would then have a choice: I could give out meaningful suffixes, like me+amazon, me+ebay, me+facebook, or meaningless hihg-entropy suffixes. But if I give out meaningful suffixes, then spammers can just try the email address I give out, replacing the portion after + that identifies them, with likely candidates instead. They are unlikely to succeed, but this is still security by obscurity. If we move to true security, picking high-entropy strings to white-list, that means I now need to manually track the relationship between the strings and the party they identify. Google should track that relationship, not me. It's a single key-value store. That's it. Associate me+ebay with squirehamboard so that I can give ebay "squirehamboard@gmail.com" and have it behave exactly as though I gave ebay "me+ebay@gmail.com" today. I think you're better than this. Be better. I have faith in you. You understand the issue. If you were working for Google you could solve it within 7 minutes (proof of concept). It's a single key-value store. I think you know very well the definition of security by obscurity. Be better than this. Be better!

newjersey 11y ago
How would I solve it without parking n number of aliases which are now forever unavailable for anyone else to use? You could do what I do and just create a separate google email address (mine literally has the word spam in it) that feeds into your primary Gmail but goes straight to all mail and is marked as read by default. It is funny when some websites reject my email address because the word spam in the address. Perhaps Google could make this process simpler by allowing autogenerated email addresses that feed into the real primary email without the user having to create a second email address. I think this would benefit Google as the user data isn't scattered around in multiple Google accounts. So for example I can request a new alias and Google would make one like ezJVoWkJFBPvMm98xhOUsi7X52l06RNblLXhhCFse3nwdkpCW7VUkIO7zgJqQDd at gmail Dot com and forward that to my real email address. If I ever respond to any email that came to that address, make it look like my email address is that string above. Now since it is just a random 63 (or longest allowed by Google) char alias that the user didn't choose, it is unlikely that it will be anything someone wants as their email address. Is this along the lines of what you were thinking?
- logicallee 11y ago
  You think extremely similar to me!!! For the sites that reject mailinator and its aliases, I give them a gmail address that literally has the word spam in it, just as yours does. But I just check it for any registration I need to confirm, I don't forward it to my main account. But it literally has the word spam in it, just like your spam gmail :) Interestingly, I've never once had any site reject that due to having the word spam in it - but maybe because it's my second-tier possibility only for use on sites that reject mailinator. I thought it would be likely that some sites would reject it for that reason, but I just haven't had that issue. What you've written: >Perhaps Google could make this process simpler by allowing autogenerated email addresses that feed into the real primary email without the user having to create a second email address. Is precisely what I suggest. Specifically, all of the current me+comment@gmail.com (which feeds this email address to the me@gmail.com email address, marking it with your comment)should be kept, it's just that instead of giving Bob's Spam Factory the email address "me+bobspam@gmail.com" I first request a high-entropy version of "me+bobspam", and get something similar to your suggestion: >ezJVoWkJFBPvMm98xhOUsi7X52l06RNblLXhhCFse3nwdkpCW7VUkIO7zgJqQDd@gmail.com and give Bob's Spam Factory that address instead. Then Bob's Spam Factory can't just take out the part after the + to get my primary email address. However, you do not need 63 random characters in order not to worry about taking possibilities from someone!!! (Not even close). Given a dictionary of 100,000 words, 3 random words have 100,000 * 100,000 * 100,000 = 1,000,000,000,000,000 or 1 quadrillion possibilities. It's vanishingly unlikely that anyone would ever pick the same strings. This is what I meant by "entropy" which is a technical term. To see this in action, consider that I came up with the "high-entropy" string squirehamboard for my example. I haven't checked yet, but it is extremely likely (99.99%) that I'm the first person to have ever written the words squireharmboard one after the other, and that if I hadn't written it, nobody would write it in the next thousand years (except exhaustively listing all possibilities, but I mean actually picking it or writing it by hand). Let's check. Let's google squirehamboard: https://www.google.com/search?q=squirehamboard https://www.google.com/search?q=squirehamboard - as you can see, nobody in the history of the Internet has written that short phrase except me. It's not the only example I had. I also listed in my original comment buffaloaerypyrite - again, it is extremely likely that I'm the only person to have ever written buffaloaerypyrite in the history of the Internet (I used a high-entropy free association method to come up with these words), and that if I hadn't written it, nobody would in the next thousand years. Let's check: https://www.google.com/search?q=buffaloaerypyrite https://www.google.com/search?q=buffaloaerypyrite Again, I'm the only person to have ever used the string buffaloaerypyrite on the entire Internet in any context. (Of course, some web server could decide to exhaustively list all three-word possibilities, since my personal mental dictionary that I free-associated through to get the terms Buffalo, Aery, Pyrite, as well as Squire, Ham and Board, from does not have 100,000 words in it - I know more like a few thousand words personally. It's possible for someone to list buffaloaerypyrite in a huge (multi-multi-terabyte) collection of passwords. But I don't quite require Google's handles to be as strong as passwords. They should just be long enough not to step on anyone's toes who are making real, normal email addresses. In practice 3 random words from a huge dictionary is more than enough. In practice you will never have any conflict with any real email address anyone chooses. But yeah, you get the basic idea. You just go overkill on the length of password that you think needs to be chosen :) 3 random words given by Google are more than enough...