Timeline for Let's Plan the Second Iteration of the Stack Exchange Quality Project!
Current License: CC BY-SA 4.0
31 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 18, 2021 at 11:34 | history | edited | CommunityBot |
replaced http://chat.stackexchange.com with https://chat.stackexchange.com
|
|
| Oct 18, 2016 at 15:02 | comment | added | ArtOfCode | @IanRingrose Not reject the post - act as if the post has been accepted, but delete it immediately after posting. Most spam systems won't go back and check. | |
| Oct 18, 2016 at 15:00 | comment | added | Ian Ringrose | @ArtOfCode so that the software posting it does not know that it has failed an automatic spam detection test, otherwise the posing software can learn to defeat the test. | |
| Oct 18, 2016 at 14:36 | comment | added | ArtOfCode | @IanRingrose If the system knows it's spam with that level of certainty, why not just delete it instead of hiding it for 5 minutes? | |
| Oct 18, 2016 at 14:35 | comment | added | Ian Ringrose | @ArtOfCode, maybe a review queue, but note I said "detected spam", e.g. the system already knows it is spam. | |
| Oct 18, 2016 at 14:33 | comment | added | ArtOfCode | @IanRingrose Then how do we see it to flag it? | |
| Oct 18, 2016 at 14:29 | comment | added | Ian Ringrose | What about not showing the detected spam to someone coming for anther IP address, and just waiting 5 minutes before the auto delete is done? | |
| Oct 18, 2016 at 9:42 | comment | added | Lundin | @Magisch Of course it might not always be feasible, it's on case-to-case basis. | |
| Oct 18, 2016 at 7:18 | comment | added | Magisch | @Lundin Do you really think some spammer sitting in russia, china, india or some obscure small country or even using hacked boxes is going to care that you report them? | |
| Oct 18, 2016 at 6:35 | comment | added | Lundin | Since SE originates from programmers, so everyone tries to solve problems like programmers: make better filters, make bots, create intricate ban systems etc etc. Maybe think outside the programmer-box instead? The only effective counter-measure long-term is to take legal action against the spammers, or at the very least report them to their ISPs. | |
| Oct 17, 2016 at 23:56 | comment | added | ArtOfCode | @Jefromi Given that we can positively identify 10 or more distinct groups of spammers, I think you're right it's not that organised :) | |
| Oct 17, 2016 at 23:55 | comment | added | Cascabel | (An alternative theory is that they're trying to circumvent the blocking they have rapid feedback on, and when they periodically succeed, that's what you pick up with Smokey. But I haven't gotten the impression it's quite that organized.) | |
| Oct 17, 2016 at 23:46 | comment | added | Cascabel | Well, so it would reject more posts immediately, but yes, I get that. My point is that rapid feedback is a real issue in some cases, and that the reason it turns out okay in this case seems to be that the spammers apparently don't care enough to respond to the feedback. It's not about whether the logic is concealed (existing SE spam detection) or open-source (Smokey). | |
| Oct 17, 2016 at 23:38 | comment | added | ArtOfCode | @Jefromi The built-in system already rejects posts immediately. The only thing that would change with integration would be that Smokey feeds the system to tell it what to reject. | |
| Oct 17, 2016 at 23:12 | comment | added | Cascabel | @ArtOfCode It's not about whether it's possible to circumvent, it's about whether they realize that they need to circumvent it, and how quickly they're able to figure out how. Rapid feedback (e.g. rejecting posts) does make both more likely. I do know a thing or two about SEO spam, and can tell you that feedback speed there is a very genuine concern. It may still not be an issue here, because those spammers are probably also spamming a ton of other places and may not notice/care too much about rejection here, but that's a different reason. | |
| Oct 17, 2016 at 20:36 | comment | added | ArtOfCode | @IanRingrose They can already figure out how to pass Smokey if they want to; the code is freely available on GitHub. They don't, though. | |
| Oct 17, 2016 at 20:22 | comment | added | Ian Ringrose | @TimPost Be careful, if you give feedback to the spammer to quickly about their post, they can learn how to bypass SmokeDetector, at present they don’t know they have been detected, as most spamming systems don’t check for results after getting conformation the post has been made. | |
| Oct 17, 2016 at 19:16 | comment | added | ArtOfCode | @dorukayhan That - but we've got ideas to solve that too. | |
| Oct 17, 2016 at 18:59 | comment | added | user8397947 | @TimPost Note that not all filters of SmokeDetector have a high true positive to false positive ratio. If we ever integrate SpamRam and SD we should only use the most successful filters (such as links following arrows in body and known-bad websites in title). | |
| Oct 17, 2016 at 18:34 | comment | added | Bohemian | @Robert I was going to start a thread on "spam". It seems some (younger?) people take the definition from the verb form "to spam" that apparently means "to pepper someone with a lot of junk" and use the flag to mark the posts as a thing used to spam the site. ie About ¾ of the "spam" flags I see are not spam, but just "annoying junk" with no "advertising" content (which I decline, usually with an explanatory custom message). Maybe there should be a custom "This is not spam! Just downvote or vote to delete" canned decline response | |
| Oct 17, 2016 at 18:08 | comment | added | Thomas Owens | On Stack Exchange, the term "spam" has, historically anyway, been limited to unsolicited commercial advertisements. A good example of that is the spam flag. This is very different than other communities and forums on the Internet, which often use the term "spam" to refer to any undesirable or unwanted content. I agree with @RobertHarvey that we should be consistent. I think favoring the narrow definition is better, since it's a particular type of flag. I've rejected spam flags unwanted posts, but not unsolicited commercial advertisements because it doesn't meet the SE definition of spam. | |
| Oct 17, 2016 at 18:01 | comment | added | user102937 | That definition appears to mean "anything unwanted." I ask, because UCE seems to be well-controlled on the Stack Exchange sites, while question quality is still somewhat in "The Wild West" territory. If we're going to talk about these things intelligently, I think it's important to make the distinction. | |
| Oct 17, 2016 at 17:59 | comment | added | ArtOfCode | @RobertHarvey I subscribe to the definition set out in the MSE FAQ on the subject | |
| Oct 17, 2016 at 17:58 | comment | added | user102937 | By "spam" do you mean "unsolicited commercial advertisements?" Or do you subscribe to the looser form of the term which basically means "anything unwanted?" | |
| Oct 17, 2016 at 17:56 | comment | added | hichris123 | @TimPost I'd much rather see other stuff prioritized rather than spam (like bullets 1 & 2). I say this even being one of those who works on SmokeDetector. :) | |
| Oct 17, 2016 at 17:54 | comment | added | user50049 | @hichris123 Well, he added a bit more context to it (while enthusiastically agreeing that we need to do more) - I think he's just pretty happy that we're getting close to getting to it :) | |
| Oct 17, 2016 at 17:45 | comment | added | hichris123 | This was already discussed in the question... "We're looking at some upgrades to the anti-spam and abuse mitigation systems. They've been good to us, but we need to keep them relevant. There are some awesome community-run projects that have been helping tremendously, and we're currently looking at ways that we might be able to integrate them more properly." | |
| Oct 17, 2016 at 17:24 | comment | added | ArtOfCode | @TimPost Yep, makes sense. We've got some ideas about how things can be interested - just yell when you want to get discussing :) | |
| Oct 17, 2016 at 17:22 | comment | added | user50049 | That's precisely what we're going to look at. I designed the original system and it has been holding, but we need to keep it modern. The problem with the system now is 100 spam posts come in, they're blocked within seconds, but there's still 100 spam posts to clean up which gives the perception of swiss cheese on the front end. Integration with Smokey at the first instance could conceivably prevent 70 of them from being posted, we just need to figure out how we can make it happen. | |
| Oct 17, 2016 at 17:20 | comment | added | Pops Staff | Yup, the reason I was looking at Smokey then was ahead of this project. Be back with you soon! | |
| Oct 17, 2016 at 17:17 | history | answered | ArtOfCode | CC BY-SA 3.0 |