If your interests include OnlyFans, pro-China propaganda, and social media accounts with artificially generated faces, then this is the botnet for you. (There's also cryptocurrency spam involved because of course there is.) #FridayShenaniGANs
This network consists of (at least) 1627 Twitter accounts created between April 30th and May 20th, 2023. All of them follow at least two of the following accounts: @AoTJewels, @BlackYellow, @Fenerbahce, and @elonmusk ,and tweet via Twitter for Android or the Twitter Web App.
The 1627 accounts in this network all use GAN-generated faces as their profile pics. All of these images have neutral backgrounds and are 255x255 pixels (Twitter default is 400x400).
(GAN = "generative adversial network", the technology used to generate the "face" images)
Unmodified StyleGAN-generated face images have the interesting property that the major facial features (particularly the eyes) are in the same position on every image. This becomes obvious when one blends multiple images together (such as the 1627 "faces" used by the network).
The accounts in this network tweet in multiple languages, with English being the most frequent (66.1% of tweets), followed by Chinese (16.1%). Slightly over half of their content is retweets (56.8%); the content retweeted is mostly NFT/cryptocurrency-related.
The network's content is repetitive, with large numbers of accounts tweeting identical or similar tweets, but varied in theme. The most common types of English tweets are similarly-worded porn tweets and famous quotes.
The network's repetitive Chinese tweets, on the other hand, are often political (although there's some porn in there too). Topics include the Australian Strategic Policy Institute, Guo Wengui, Steve Bannon, and Yan Li-Meng, all of whom the network's tweets portray as anti-China.
Several of the accounts in this network also recently tweeted propaganda about Moldova. More info in this thread and blog post (H/T @z3dster). medium.com/@mitchchaiet/m…
Tips on detecting StyleGAN-generated faces (note that some of these do not apply to output from text-to-image models such as Stable Diffusion or Midjourney):
Meet @thisisorange, a Twitter account created in February 2022 with a gold "verified organization" badge, thousands of batch-created fake followers, and a couple other interesting traits.
Verified organizations on Twitter can verify affiliated accounts (employees, teams, brand names, etc), which receive blue checkmarks as well as an organization badge (help.twitter.com/en/using-twitt…). The @thisisorange account has thousands of affiliates, mostly cryptocurrency accounts.
How did this come about? The website linked on @thisisorange's profile (orange dot associates) apparently allows one to become an affiliate simply by providing a Twitter account and a cryptocurrency wallet.
It's a Saturday in May, and a reply spam network is busy spamming people's replies with extremely similar tweets encouraging them to log into an obscure website with supplied login credentials. (The login info is the same in every reply.)
The website promoted by the spam accounts (uu55111 dot com) describes itself as a "financial income platform" and claims to offer various crypto-related financial services. Needless to say, obscure financial websites promoted by spam networks should be approached with distrust.
The reply spam network promoting this website consists of (at least) 9688 accounts created in March/April 2023. All have usernames ending in random digits and all have thus far tweeted exclusively via the Twitter Web App.
It's a great day to look at a network of mutually interacting Twitter accounts that use TweetDeck to amplify one another's content, most (but not all) of which is Florida-related. #FloridaSpam
Much of the network's content consists of brief tweetstorms where several of the accounts in the network tweet or retweet the same thing a few seconds or minutes apart. Overall, approximately 70% of the network's tweets/retweets were duplicated by at least 3 of the 7 accounts.
Popular chatbot ChatGPT will sometimes return error messages when asked to produce offensive content. If you're a spam network operator who uses ChatGPT to generate spam and you're not paying attention, these messages will show up in the spam you generate.
This spam network consists of (at least) 59645 Twitter accounts, mostly created between 2010 and 2016. All of their recent tweets were sent via the Twitter Web App. Some accounts have old unrelated tweets followed by a multi-year gap, which suggests they were hijacked/purchased.
Although most of this network's tweets are (superficially) unique, there are some exact duplicates, which fall into two categories:
If you're someone with integrity at @thedailybeast, this is a good time to give the public a look at what's going on behind the scenes and why an article containing demonstrable falsehoods and harassment from a known cyberstalker is still on your website.
Given the amount of time that has passed, the degree of pushback, and the repeated debunking of the article's primary source and elements, there is no scenario in which this is innocent. Someone at The Daily Beast is 100% aware the article is false and is keeping it up anyway.
This isn't going away, and the damage to the reputations of The Daily Beast and everyone associated with it grows harder to reverse with each passing moment. In life, there are bells you don't unring, and platforming an abusive cyberstalker is one of them.
The author of this article, @JakeLahut, did not reach out to me before publishing false claims that I participate in "online harassment campaigns". If he had, I would have told him the claims were lies that originated with a pair of pathological liars.
The claim at the beginning of this article about an "altered image" of cyberstalker Steven Jarvis's child is also a lie. The image in question was a GAN-generated face taken from ThisPersonDoesNotExist, and was not an "altered image" of any actual human.