User talk:Praxidicae/fakenews

From WikiProjectMed
Jump to navigation Jump to search

@Billinghurst:, @علاء:, @Vexations: since you all have been helping so much with this, I'd like to point you guys to User:Praxidicae/fakenews#"duleweboffice", which I think has more than enough evidence to immediately blacklist (preferably globally) and will need some cleanup as I suspect for at least 2 sources they were at one point legitimate but they sniped the domain and now it's all spam. Also including Alaa as a steward and ar admin since there is a lot of this spam there as well. Praxidicae (talk) 14:04, 21 January 2020 (UTC)[reply]


And for anyone questioning why these are listed here, please take a look at this about us page and take note that "Joel Woodley" is actually a pro hockey player, Joël Vermin, "Keith Hoffman" is pro-hockey player Keith Kincaid and in the interest of not outing the rest of the people, they certainly don't work for that site and are easily found but low profile. Praxidicae (talk) 14:50, 21 January 2020 (UTC)[reply]

 Comment: hilarious as you actually look at the Joel image in full, you see the jersey ... https://tineye.com/search/27495a4c71d56602a93eb3ac580eca669caa1db5?page=1

Also finding some of the global-search results is bringing up some good inter-relationships (example

This is going to be ongoing as they extend deepfakespam farming, and I also think that this stuff will further evolve. So whilst starting the build here works, I am not certain that it is sustainable into the future, especially as we expand investigations and analysis. Some of the templating at Meta is more readable (to my eyes) and suitable for this tracking, especially with its reverse linking, so some clarity there would be helpful.

@Beetstra: I think we need to alert you to this stuff as it will involve COIBot reports and how those compilations may take place. — billinghurst sDrewth 23:12, 21 January 2020 (UTC)

@Billinghurst: Fine with that. I've upped the maxrecord setting for COIBot reports to 1000 (last edit to meta settings; countlimit) .. I would not standard go higher, but you could put it to 5000 or even more for a couple of hours while COIBot churns out some reports (note that COIBot may have to parse 4999 edits which may make the report make a very long time to save and that the LinkSaver module may crash on such amounts of data - that is why I limited it on the previous box - this box is bigger). Those reports will need to be 'split' manually then, as they will likely overflow the parser limits: put subsection headers after each 100-or-so links, save, and then when you 'edit' a subsection that will be fully visible (I should build this into COIBot -> ).
For ease of working, COIBot can be 'poked' from other pages by adding pages to the settings (you could add 'User:Praxidicae/fakenews@en.wikipedia.org' to the meta-settings (pokepages, any linksummary and UserSummary there will then get poked and a report generated by COIBot - note that coibot likely needs a restart as that is built into the core of the bot and it will not re-read the settings otherwise).
  • pokepages are just scanned for added LinkSummary and UserSummary templates when added by grantedusers
  • autopokepages are scanned for added LinkSummary and UserSummary templates by any user and a report will be created, those are our local and global discussion pages.
  • linksummarypages when LinkSummaries are being posted here, COIBot will from then monitor the link (it shows something on IRC, and you will see a 'domainredlist'-statement in the report, like in m:User:COIBot/UserReports/50.239.116.194).
--Dirk Beetstra T C 05:30, 22 January 2020 (UTC)[reply]
@JzG: who has a sweet spot for fake news :-) .. --Dirk Beetstra T C 05:36, 22 January 2020 (UTC)[reply]
I’m mobile at the moment but I have a list set up for meta and am tempted to do it on a private wiki with imported templates to avoid giving away all the tells. Praxidicae (talk) 23:29, 21 January 2020 (UTC)[reply]

another example

taken from My Country Mobile, where an IP from Rajasthan cites http://brandnewstoday.people.msnbc.com/_news/2020/01/21/38814887-my-country-mobile-offers-voip-for-businesses-around-the-world-voip-for-business] which could easily be mistaken by an unsuspecting reader for editorial content. Newsguard [1] gives it a cautiously positive rating. The profile photo is a stock photo [2] and almost certainly not a photo of "Julia Rieves"; their "Journalist and Correspondent based in San Francisco" What they really post is content from brandnewstoday.com presented as if it were from msnbc.com. The contact page is telling: It's apparently run by Newspaper Comunication [sic] Service, and tells us that "You are a part of the Rebel Alliance and a traitor so what do you think of her Han." Vexations (talk) 04:23, 22 January 2020 (UTC)[reply]

To see another example how such citations are used, have a look at https://en.wikipedia.org/wiki/Natalie_Lefevre#cite_ref-6, which claims "n 2018, she was chosen as The Editor's Pick of the Best New Influencers at MSNBC". That article is actually from kivodaily.com, promoted by a fake "Tyler Penske" with a profile photo from a stock agency. Vexations (talk) 04:40, 22 January 2020 (UTC)[reply]
Vexations, that article needs nuking. The only sources are churnalism. Guy (help!) 09:04, 22 January 2020 (UTC)[reply]
(ec) @Vexations: kicked some reports which will appear at m:User:COIBot/XWiki/brandnewstoday.people.msnbc.com and m:User:COIBot/XWiki/brandnewstoday.com and some of the links may also have value. Also has global-search results. — billinghurst sDrewth 09:05, 22 January 2020 (UTC)[reply]