A recent investigation by Dismislab has uncovered a sophisticated political bot network operating on Facebook, which involved 1,369 fake accounts posting over 21,000 coordinated comments in favor of Bangladesh’s former ruling party Awami League.
The network, active around the time of the 12th parliamentary election, continued its activities long after the election had concluded.
The discovery began on June 21 with an article titled, “Is there an invisible code on every page printed by a color printer?” published by the online news outlet bdnews24.com.
The article itself was unrelated to politics, but a comment beneath the article’s Facebook post caught the attention of a Dismislab researcher.
The comment, which claimed that the “upcoming election” would be fair and accused the opposition party, Bangladesh Nationalist Party (BNP), of fearing participation, sparked curiosity due to its political nature and timing.
Further investigation revealed that even months after the election, Facebook users were still posting comments critical of the opposition party on non-political posts.
This unusual activity prompted a deeper analysis, leading to the identification of a bot network responsible for thousands of similar comments across various Facebook pages.
The bot network, which targeted posts related to political content or containing specific keywords, demonstrated several errors in its execution.
For example, the network mistakenly identified unrelated terms containing the letters “EC” (Election Commission) in Bangla script and posted political comments on irrelevant topics.
This mistake led to comments being placed under news articles unrelated to politics, such as stories about a bank robbery or a medical incident.
The research team compiled a database of around 35,000 comments, identifying over 21,000 as being posted by the bot accounts.
These accounts showed typical traits of political bots: locked or private profiles, minimal personal information, and coordinated behavior across multiple pages.
The study revealed that many profiles were created just before the election and exhibited patterns such as using stolen profile pictures, having a low friend count, and following specific pages affiliated with the ruling party.
Meta, the parent company of Facebook, defines Coordinated Inauthentic Behavior (CIB) as coordinated efforts to manipulate public discourse using fake accounts.
The activities of this bot network appear to align with such behaviour, using computational tools to post pre-written comments on selected posts.
Dr. Naeemul Hassan, an assistant professor at the Philip Merrill College of Journalism and Information Studies at the University of Maryland, told Dismislab: “The findings of the investigation suggest the use of computational tools.”
However, he emphasized that while the coordinated nature of the comments suggests computational involvement, the investigation lacks direct evidence of full automation.
The network predominantly targeted 42 Facebook pages, including media outlets and pages associated with the BNP.
The content of the comments was primarily critical of the BNP and its leaders or supportive of the Awami League and the Election Commission.
The discovery of this bot network raises significant concerns about the use of computational propaganda and its potential impact on democratic processes.
Dr Hassan warned that such technologies are likely to become more advanced with the rise of artificial intelligence, making it even harder to detect coordinated misinformation campaigns in the future.
Profile patterns
Dismislab's investigation found that a significant majority of these fake profiles (77%) were under female names, with the name "Akter" frequently appearing as a surname in 24% of these cases.
Male profiles often used surnames like "Ahmed" and "Islam."
Interestingly, 90% of both male and female profiles consisted of two-word names, sometimes created by splitting a single name, such as "Ri Pa" or "Mi Na."
The profiles displayed several traits typical of fake accounts as defined by fact-checking organizations and cybersecurity firms such as Snopes, Cloudflare, and McAfee.
These traits included restricted privacy settings, minimal personal information, and profile pictures sourced from elsewhere on the internet.
Of the 1,369 profiles analyzed, 247 were completely locked, and 70% of the remaining profiles used profile pictures that were found to be taken from the internet, sometimes with the same picture used across multiple accounts.
For example, a particular profile picture that could not be found elsewhere online was used simultaneously by eight different profiles.
This picture was uploaded on the same day, November 30, 2023, by accounts named Fahad Islam, Rajib Ahmed, Mamun Ahmed, Ahmed Tahsin, Riyad Khan, Tanzim Ahmed, Jahangir Alom, and Ariyan Islam.
This trend of using the same picture across multiple accounts was noted in 19 other instances, involving 111 accounts in total.
The investigation also found that 85% of the unlocked profiles lacked any identity information in the "About" section, and 93% had no public posts except for profile pictures and cover photos.
Nearly half of these profiles (49%) did not display friend counts, and 45% had fewer than 200 friends.
Only seven accounts were found to post regularly, typically sharing similar types of content.
The research identified 21,221 political comments made by these fake profiles, but only 474 of these comments were unique.
The comments were frequently recycled across various posts, indicating a high level of coordination among the bots.
For instance, a profile named Riya Akter made 138 comments across different posts over six months, including a comment about the fairness of the upcoming election, which was posted 96 times by 109 different bot accounts.
Timing and targeting flaws
The analysis also uncovered flaws in the bot network’s operations, with 88% of the comments posted within a narrow time window of two hours.
On one occasion, 151 comments were made on a news link about BNP Chairperson Khaleda Zia's treatment between 11:45am and 1:40pm.
Furthermore, the network displayed poor targeting, posting political comments on irrelevant news items.
For example, comments criticizing the opposition or praising the government appeared under posts about a bank robbery, an obituary for former Iranian President Ibrahim Raisi, and even a cricket match venue change request by the ICC.
The errors extended to outdated references. Bots continued to use phrases like “the upcoming 12th National Parliamentary Election” long after the election had occurred.
Comments created before the election were posted indiscriminately afterward, indicating a lack of human oversight in the operation.
It may seem puzzling at first why these political comments appear under unrelated posts.
However, a closer analysis of the text accompanying these posts reveals a common pattern.
When certain terms - like ICU, ICC, Raisi, IFIC, OIC, ICT, Dialysis, or Crisis - are written in Bangla script, they contain the sequence of letters that correspond to “EC” in Bangla.
For instance, the bdnews24.com post about the color printer, which sparked this research, included the term “MIC” (Machine Identification Code), which also contains this letter sequence when transliterated into Bangla.
“EC” is an abbreviation for the Election Commission.
The bot network jumped on any post containing the term “EC,” whether relevant or not, and posted coordinated political comments there.
This suggests that the network used computational tools based on keyword detection to select which posts to comment on.
In 2022, Meta identified a Russian CIB network where similar bot-comment descriptions were found under irrelevant posts.
The social media company also mentioned a similar Israeli network in its 2024 Adversarial Threat Report for the first quarter.



