Fighting Propaganda With Technology — Part 1
Terms like inauthentic cyber activity, bot networks, and disinformation are frequently heard when we log on to social media, amidst the usual “influencer stuff.” However, these topics are rarely discussed in-depth on 8pm talk shows. In this article, we will provide a detailed explanation which will also help you recognize these phenomena with the technology available to you. We will walk you through one of the recent cases we worked on with the Network Contagion Research Institute (NCRI), a US-based organization.
We investigated the Khalistan bot network that were spreading fake news and creating havoc in the Indian cyberspace, while Khalistani radicals carried out physical attacks on Indian embassies. This is just one example of the many bot networks causing havoc, but with proper analysis of their behavior, earlier warnings could have been issued and some attacks could have been avoided.
During the Khalistan event, one name that gained prominence in India was ‘Amritpal Singh.’ We started our investigation by searching for tweets that used the hashtag #WestandwithAmritpalsingh, and using our tool, we obtained the following information:
The density of account creation during a three-month period and the tweet volume for Amritpal Singh.
(Note: Density here has a maximum value of 1 when summed together, so a peak of 0.9 means 90% of the total over the period.)
The peak for account creation is between 16/3/2023 to 18/3/2023, and the peak of the tweet volume is on 19/3/2023. Total of 1750 accounts were created during the period of 16–19 march.
How to Analyze this:
1. Check the tweet timeline on the Hashtag
2. Check the creation date of accounts posting the hashtag
If they coincide as in the case above, then it is an indication of “Propaganda”
We then scrapped the tweets of the account and did various analyses that included the heat map to show the duration of the day in which the accounts were most active. Type of tweets they did i.e., textual & non textual. Then we used the cosine similarity method to find similarity between accounts. This will be explained in detail in part 2 of the article.
We looked at the number of tweets from each account and not surprisingly, almost 47% of the accounts have not even done 50 tweets overall, this number is inline because almost all of the 1750 accounts created between 16–19 March 2023, fell under this category and i guess they were yet to start their heavy trolling before we caught them.
Our tool also filtered out other hashtags that were trending alongside #WestandwithAmritpalsingh. Take a look:
The NCRI report:
Our collaborators at NCRI found a network that was heavily posting images and tweets in support of Khalistan. These accounts were posting 25–30 tweets at once within a few minutes, often repeating the same tweet multiple times, as if they were running through a Python code with a for loop “for i in range(30)”.
Although it was clear that these were bots, the challenge was to identify the driver of the network, and we used multiple methods to achieve this.
1. We found accounts with the highest number of followers in the network because we saw these accounts were following each other as well. We then scrapped the bio of all the followers of this account and here it was a clear picture of who was driving it.
Any guesses?
Yupp…right. PTI Pakistan. But we wanted to get more evidence.
2. We scrapped all followers of followers of one of the main bot accounts and made a network graph. This clearly highlighted all the PTI drivers at the end with big circles.
You can read more about the finding of the 2nd bot network (one driven by PTI in NCRI’s report)
Our findings and suggestions
The attacks on social media have become sophisticated but with our more sophisticated technology it is not difficult to pre-empt and stop these attacks. We say people lie sometimes but I feel the tech doesn’t so I have coined a term which I call the “#TechTruth”. The proof we collect through our tech and tools are enough to reveal the people driving such networks aka the “bad guys”.
If we act in time we can track these social bullies and their posts can even help us predict their actions on ground.
Rohit
Founder & Director (Thinkfi.net)
Contact: Thinkfi.net@gmail.com
