This pornbot network has grown substantially since we last looked at it. 111 of the accounts we previously found have been removed by Twitter, but 1421 new ones have joined the chorus, for a total of 1813 accounts.
The output of this pornbot network continues to grow as more accounts are added to it, peaking at over 300 tweets per hour. Almost all tweets are (hypothetically, at least) posted via "Twitter Web Client", the old and no longer available version of the Twitter website.
The accounts in this pornbot network all have 8-character names composed of random numbers and lowercase letters. All have liked zero tweets and follow zero accounts. Tweets are in Japanese and encourage users to DM the bots.
As before, nearly all of the accounts in this pornbot network have anime/manga profile pics. We decided to show them in video form since a collage of 1813 images is a bit difficult to look at.
Another data point suggesting that these accounts are part of the same operation (not that we're in much doubt given the other similarities): almost all tweets from this botnet were processed by the same Twitter datacenter.
(Background on how to discern datacenter from tweet ID. We've noticed that sometime around the beginning of 2020 the set of datacenter IDs in use changed, so take that into account if you choose to make use of this technique yourself.)
The first recent #DontBelieveArmenia tweet (sporadic earlier occurrences exist) was sent yesterday (Oct 5 2020) by @CoronaMan19, an account created in March 2020 with 21 tweets and 1 follower. It spiked early this morning (Oct 6 2020), frequently exceeding 200 tweets per minute.
Although the accounts involved in the #DontBelieveArmenia trend aren't provably automated (based on apps/24 hour activity), the content is incredibly repetitive. The repeated tweets include spam directed at @iamcardib, as noted earlier by @josh_emerson.
In the wake of Donald Trump's positive COVID-19 test and subsequent hospitalization, #MAGA Twitter has been abuzz with conspiracy theories attributing his diagnosis to an assassination attempt by various entities such as China, Biden, Pelosi, and Alexei Navalny.
We downloaded tweets containing the word "assassination" subsequent to Trump's announcement of his positive test, and filtered out some obvious false positives (tweets about the Jamal Khashoggi assassination, for example). Very little of the traffic appears to be automated.
Tweets theorizing that Trump's COVID-19 diagnosis was actually part of an assassination attempt began appearing a scant 3 minutes after his positive test was announced (Oct 1 2020), starting with a reply from 4-month old account @Freddie111112 to CNN reporter @mkraju.
If you're looking for an automated military patriot mom who is sick of politics as usual and spams links to Gateway Pundit articles via IFTTT, @PatriotMomNDJ just might be the Twitter account for you. #SaturdaySpam
All of @PatriotMomNDJ's recent tweets are sent via IFTTT, but this account has utilized other Twitter automation tools in the past, mainly SocialOomph. Almost all of its original tweets are automated (140161 of 140110, 99.96%), so this account is definitively a bot.
Similarly, @PatriotMomNDJ currently only links Gateway Pundit, but in the past it varied its spam lineup. Right-wing news site The Blaze was a frequent flier until late 2018, and commercial spam (Amazon, staged(dot)com, and others) was a major portion of its early content.
We've done a decent amount of digging into various bulk follow botnets over the last couple of years, and decided to aggregate all of our threads on the topic. Some botnets are still online, while others have been shut down by Twitter.
It's October 1st, which is as good a day as any to look at yet another fake follower botnet. This network consists of (at least) 95848 accounts created between December 25th, 2016 and December 31st, 2017.
The accounts in this botnet have much in common. All have account names consisting of random lowercase letters followed by 1-4 digits, Japanese display names, few or no followers, few or no likes, and tweet(ed) exclusively via the Twitter Web Client. Most tweets are in Japanese.
How did we find the set of accounts in the network? We noticed ~20000 of them following @Arslan_Sadiq, and mapped out the rest of the botnet by checking the followers of the other accounts they followed for additional accounts that fit the pattern.
The first debate between Joe Biden and Donald Trump took place last night (September 29th, 2020). We took a look (or rather a listen) to determine what topics were the most loudly discussed, and by whom.
We analyzed the debate audio and identified the 50 loudest peaks, based on a rolling one-second average of the amplitude of the audio signal. The loud portions became increasingly frequent as the debate went on.
Next, we listened to the 50 loudest peaks, along with the surrounding audio for context. In multiple cases, multiple peaks were only a few seconds apart and were part of the same sentence or exchange, so we combined those.