Archive for September, 2017

Twitter Must Make it Easy to Mass-Report Spam Bots

I found a network of Russian female bots. Twitter spam bots.

They are not actually female. They just have Russian female names and female photos.

Most of those that I found were created in September 2016, although some were created at other times.

They all have similar taglines:

  • “In my opinion, everything is wonderful. I wonder what else” (“По-моему всё прекрасно. Интересно что ещё”)
  • “Right now absolutely everything is excellent. I wonder how else” (“Сейчас вообще всё отлично. Интересно как там ещё”)
  • “It looks like absolutely everything is wonderful. I’ll see what will happen next” (“Вроде вообще всё прекрасно. Посмотрю что будет дальше”)

… And so forth, with minor variations, which are very easy to detect for a human who knows Russian, although I’m less sure about software. (This reminds me of how I was interviewed for several natural language processing positions around 2011. All of them were about optimizing site text for Google ads, and all of them specifically targeted only English. When you only target English, other languages are used to spam you.)

Their usernames are all almost random and end with two digits: flowoghub90, viotrondo86, chirowsga88 (although “90” seem to be the most frequent digits). As location, they all indicate one of the large cities of Russia: Moscow, Krasnoyarsk, Perm, Saint-Petersburg, Rostov-on-Don, etc.

All of them post nothing but retweets of other accounts popular in Russia:

Curiously, all their names are only typical to ethnic Russians. Names of real women from Russia would be much more varied—there would be a lot of typical Armenian, Ukrainian, Jewish, Georgian, and Tatar names that reflect Russia’s diversity: Melikyan, Petrenko, Rivkind, Gamkrelidze, Khamitova. But these spam bot accounts only have names such as Kuznetsova, Romanova, Ershova, Medvedeva, Kiseleva. If you aren’t familiar with the Russian culture, let me make a comparison to the U.S.: It’s like having a lot of people named Smith, Harris, Anderson, and Roberts, and nobody named Gonzalez, Khan, O’Connor, Rosenberg, or Kim. Maybe the spammers wanted to be more mainstream than mainstream, and maybe it is just overt racism.

I found them when I noticed that a lot of unfamiliar accounts with Russian female names were retweeting something by Pavel Durov in which I was mentioned. Durov is the founder of VK and Telegram, and I guess that he can be classified under “major internet businesses” in the list above. I noticed the similar taglines of the “women”, and immediately understood they are all spam bots.

These accounts are active. Some of them retweeted stuff while I was writing this post. I also keep getting retweet notifications, more than two weeks after Durov’s original tweet was posted.

When I am looking at any of these accounts, Twitter suggests me similar ones, and they are all in the same network: Russian female names, similar “everything is wonderful” taglines, similar content. So Twitter’s software understands that they are similar, but doesn’t understand that they are spam bots that should be utterly banned. I also noticed that some of them are still suggested to me after I blocked them, which goes against the whole point of blocking.

I don’t know how many there are of them in this network. Likely thousands. I reported thirty or so, and I wonder whether it’s efficient for anything.

I also don’t know what is their purpose. Boost the popularity of other Russian accounts? But those that they retweet are popular already. Waste the time of people who try to use Twitter productively? Maybe; at least it’s the effect in my case. Function as bot followers in “pay to follow” networks? Possibly, but they have existed for a year, and they don’t follow so many people.

I’m probably not discovering anything very new in this post. But especially if I don’t, it all the more makes me wonder why isn’t this problem already addressed somehow. At the very least it should be possible to report them more efficiently with one click or tap. And Twitter should also provide a form for mass-reporting; currently, Twitter’s guides about spam only suggest this: “The most effective way to report spam is to go directly to the offending account profile, click the drop-down menu in the upper right corner, and select “report account as spam” from the list.” It’s OK for one account, but it requires five clicks, and it doesn’t scale for something as systematic as what I am describing in this post.

I do hope that somebody from Twitter will read this and do something about it. This is obvious systematic abuse, and I have no better way to report it.