DISARMframeworks/generated_pages/techniques/T0145.002.md

Technique T0145.002: AI-Generated Account Imagery

• Summary: AI Generated images used in account imagery.
  
  An influence operation might flesh out its account by uploading account imagery (e.g. a profile picture), increasing its perceived legitimacy. By using an AI-generated picture for this purpose, they are able to present themselves as a real person without compromising their own identity, or risking detection by taking a real person’s existing profile picture.
  
  Associated Techniques and Sub-techniques
  T0086.002: Develop AI-Generated Images (Deepfakes): Analysts should use this sub-technique to document use of AI generated imagery used to support narratives.

• Belongs to tactic stage: TA15

Incident	Descriptions given for this incident
I00082 Meta’s November 2021 Adversarial Threat Report	“[Meta] removed 41 Facebook accounts, five Groups, and four Instagram accounts for violating our policy against coordinated inauthentic behavior. This activity originated in Belarus and primarily targeted audiences in the Middle East and Europe. “The core of this activity began in October 2021, with some accounts created as recently as mid-November. The people behind it used newly-created fake accounts — many of which were detected and disabled by our automated systems soon after creation — to pose as journalists and activists from the European Union, particularly Poland and Lithuania. Some of the accounts used profile photos likely generated using artificial intelligence techniques like generative adversarial networks (GAN). These fictitious personas posted criticism of Poland in English, Polish, and Kurdish, including pictures and videos about Polish border guards allegedly violating migrants’ rights, and compared Poland’s treatment of migrants against other countries’. They also posted to Groups focused on the welfare of migrants in Europe. A few accounts posted in Russian about relations between Belarus and the Baltic States.” This example shows how accounts identified as participating in coordinated inauthentic behaviour were presenting themselves as journalists and activists while spreading operation narratives (T0097.102: Journalist Persona, T0097.103: Activist Persona). Additionally, analysts at Meta identified accounts which were participating in coordinated inauthentic behaviour that had likely used AI-Generated images as their profile pictures (T0145.002: AI-Generated Account Imagery).
I00088 Much Ado About ‘Somethings’ - China-Linked Influence Operation Endures Despite Takedown	“The broader War of Somethings (WoS) network, so dubbed because all the Facebook pages and user accounts in the network are connected to “The War of Somethings” page,  behaves very similarly to previous Spamouflage campaigns. “Spamouflage is a coordinated inatuhentic behaviour network attributed to the Chinese state. “Despite the WoS network’s relative sophistication, there are tell-tale signs that it is an influence operation. Several user profile photos display signs of AI generation or do not match the profile’s listed gender.” A network of accounts connected to the facebook page “The War of Somethings” used AI-generated images of people as their profile picture (T0145.002: AI-Generated Account Imagery).
I00091 Facebook uncovers Chinese network behind fake expert	“Earlier in July [2021], an account posing as a Swiss biologist called Wilson Edwards had made statements on Facebook and Twitter that the United States was applying pressure on the World Health Organization scientists who were studying the origins of Covid-19 in an attempt to blame the virus on China. “State media outlets, including CGTN, Shanghai Daily and Global Times, had cited the so-called biologist based on his Facebook profile. “However, the Swiss embassy said in August that the person likely did not exist, as the Facebook account was opened only two weeks prior to its first post and only had three friends. “It added "there was no registry of a Swiss citizen with the name "Wilson Edwards" and no academic articles under the name", and urged Chinese media outlets to take down any mention of him. [...] “It also said that his profile photo also appeared to have been generated using machine-learning capabilities.” In this example an account created on Facebook presented itself as a Swiss biologist to present a narrative related to COVID-19 (T0143.002: Fabricated Persona, T0097.106: Researcher Persona). It used an AI-Generated profile picture to disguise itself (T0145.002: AI-Generated Account Imagery).

Counters	Response types

DO NOT EDIT ABOVE THIS LINE - PLEASE ADD NOTES BELOW