Cyabra Launches Deepfake Detection

Vaccine Disinformation: Part One

Download the full report now

With COVID-19 vaccines currently being developed and deployed worldwide, the public now faces its next public health challenge- that of disinformation.

In efforts to shed light on some of the ongoing fake campaigns circulating on social media, Cyabra used its AI based solution to scan these conversations to gain a deeper understanding of these campaigns. Focusing on Facebook and Twitter, Cyabra analyzed the online behavior, connections, and messaging of 132,000 profiles. Here, we found nearly 18,000 fake profiles (13.5%). Based on Cyabra’s experience with disinformation campaigns, this percentage of fake profiles indicates the presence of an online disinformation campaign. Cyabra typically encounters around seven to ten percent of fake profiles.

Findings

Cyabra’s tools analyze the connections between each profile in order to understand the impact and reach of each profile. The image below is a cluster from Cyabra’s dashboard representing the main profiles participating in the vaccines discourse on Twitter and the manner in which they are connected. The red nodes represent fake profiles; the green nodes represent real profiles. The bigger the node, the more connections the profile has.

Network scatterplot of 2020 vaccine conversation with a dense cluster of green nodes and scattered red nodes indicating potential disinformation accounts

While the visual link analysis depicted above shows all of the profiles that interacted with one another (following, replying, or retweeting), the images below represent segmented profiles, otherwise known as “communities.” Cyabra’s “community” function highlights profiles that are highly engaged with each other and share the greatest number of connections and, often, a common theme. The themes that Cyabra uncovered relating to COVID-19 vaccines are presented in the images below. Analyzing all of the fake profiles, Cyabra discovered three Twitter communities comprised of fake profiles tweeting three distinctive sets of messaging. Two of these fake campaigns actively spread favorable tweets about AstraZeneca’s vaccine while criticizing other companies developing vaccines. The third fake campaign disputes the existence of COVID-19 and attacks the utility of all COVID-19 vaccines.

Network graph depicting social-media conversation clusters about the AstraZeneca vaccine, with varying-size red circles representing pro- and anti-vaccine groups connected by blue relationship lines

Community A: Anti-Vaccine

The anti-vaccine community contains 136 fake profiles that are actively spreading negative content against all COVID-19 vaccines.

Collage of anti-COVID-19 vaccine tweets alongside a network graph labeled Community A illustrating fake profiles spreading vaccine misinformation

Community B: Pro-AstraZeneca (1)

Community B contains 239 fake profiles circulating positive content about AstraZeneca’s vaccine progress and positive content about the company.

Visualization of a pro-AstraZeneca fake Twitter community: large orange circle with many small nodes alongside example tweets praising the vaccine’s effectiveness

Community C: Pro-AstraZeneca (2)

Community C also praises AstraZeneca but spreads harmful content surrounding Pfizer’s COVID-19 vaccine. This community contains 220 fake profiles.

Network visualization highlighting Community C, a pro-AstraZeneca fake Twitter community, surrounded by sample tweets praising AstraZeneca’s COVID-19 vaccine and no-profit pricing approach

Cyabra did not find a significant difference on Facebook between topics discussed by real profiles and the ones discussed by fake profiles. However, a trending topic that stood out amongst fake Facebook participating in the COVID-19 vaccine discourse was Bitcoin. The image below represents an example of the system’s classification of subjects, with subjects used by real profiles shown in green and ones used by fake profiles shown in red.

Side-by-side table showing the most frequently mentioned fake versus real subjects in the 2020 vaccine disinformation discussion, with topics such as Astrazeneca, Pfizer, Covid, and Russia along with their frequency and profile counts
Figure 7 – Cyabra’s topics division into real and fake profiles from the system 

While the real profiles did not discuss anything relating to Bitcoin, fake profiles used the subject of Bitcoin numerous times. Cyabra analysts found that these fake profiles spread content on Bitcoin for advertising as a part of a fake campaign.

Below is an image from Cyabra’s platform showing the connections between the fake profiles on Facebook that posted about Bitcoin in discussions relating to COVID-19 vaccines. The fake campaign is taking advantage of the online interest in COVID-19 vaccines to promote a Bitcoin website. The fake profiles identified posts related to COVID-19 vaccines with high engagement and replied to them with promotional content about Bitcoin sites and Telegram groups.

Infographic featuring a network diagram titled Bitcoin alongside screenshots of Facebook posts and comments that promote false claims about COVID-19 vaccines and Bitcoin investments, including a WHO update hijacked by spam and a Telegram invitation

Aside from communities, there is also merit to doing a deep dive of the most influential profiles. In a disinformation campaign, there are typically three types of profiles with the highest impact in a fake campaign: 1.The most content: The more posts, replies and shares a profile creates, the more influence it has in shaping the conversation, both within the campaign and with profiles that are only partially connected to the campaign. 2.The most connections: The more connections a profile has, the more it can control what people within the campaign see. 3.The most engaged: Profiles that are the most active in a campaign can shape the way people who are new to the subject perceive it.

Cyabra marks the connections between fake and real profiles to emphasize which fake accounts “break through” the fake profile sphere and can influence real profiles. Understanding which fake profiles have the highest number of real connections is another method to understand which fake profile has the most influence.

The most connected fake profile on Twitter is esme_hornbeam, with 63 fake connections. The system labeled the profile as fake due to a large percentage of its content being retweets, and its bot-oriented behavior. To identify any content possibly linked to the subject, Cyabra extracted multiple tweets that the profile tweeted about the COVID-19 vaccines.

Network visualization highlighting a fake Twitter account named esme_hornbeam with 63 connections
Figure 9.1 – The most connected fake profile
Twitter profile with a flower avatar, follower and following counts, and a bio referencing humanism, media bias, Brexit, and the NHS

Conclusion

With COVID-19 vaccines currently being developed and deployed worldwide, the public now faces its next public health challenge- that of disinformation. In efforts to shed light on some of the ongoing fake campaigns circulating on social media, Cyabra scanned 132,000 accounts where advanced AI uncovered multiple, harmful agendas. Within this sample, 18,000 profiles were fake, resulting in a significant reach of each of these disinformation campaigns. With two of the fake Twitter campaigns favoring the AstraZeneca COVID-19 vaccine, the misleading agenda intended to promote the vaccine as superior to the other COVID-19 vaccines in development. The third Twitter disinformation campaign, and perhaps the most dangerous, attacked the existence of the Coronavirus and all COVID-19 vaccines, claiming that the Coronavirus does not exist and is a guise by governments planning to control human actions. Additionally, Cyabra identified another campaign of fake profiles on Facebook that exploited the online interest in COVID-19 to advertise Bitcoin.

As disinformation becomes a more prevalent threat, Cyabra continues to monitor the disinformation surrounding COVID-19.

Related posts

#BoycottWoolworths Enhanced by Fake Profiles

Woolworths Supermarkets, AKA “Woolies”, is Australia's biggest supermarket chain, sometimes referred to as “the Aussie Walmart”. Founded in 1924, Woolworths holds a market share of...

Map of Australia overlaid with #BoycottWoolworths, #BoycottWoolies and #WokeWorth slogans, red location pins marked as Woolworths stores, and angry face and thumbs-down icons showing nationwide boycott sentiment

Rotem Baruchin

January 23, 2024

Companies, Watch Social Media For Event Risks

Any company, especially large corporations, needs to be aware of negative online conversations. This is especially true when these conversations manifest in the physical world...

Crowd of demonstrators with stop and strike signs emerging from a computer screen, representing social media protests and event risk monitoring

Rotem Baruchin

August 31, 2023

Bots Hitch a Ride on Bolt’s PR Crisis

At the end of January, ride-hailing company Bolt faced a PR crisis with a lot of the conversation playing out on X (Twitter). As Bolt...

Hand holding a smartphone displaying a city map with location pins, overlaid with network graphics connecting multiple small autonomous delivery robots against an urban street background

Rotem Baruchin

February 5, 2024