We have a long discussed the possible use of social media for both intelligence gathering and PSYOPs operations. In 2013 I wrote an interesting post about Social Media use in the Military Sector, I also defined the term “Social Network Poisoning” on Wikipedia, but while the Italian translation was still online the English one was deleted and I was banned from Wikipedia because I was trying to explain the phenomena in the last yeas made the headlines. No matter.
In the past months, security experts have collected evidence that links the Russian Government to massive propaganda operations through social networks aimed to influence the sentiment of Americans on the 2016 Presidential elections.
I was analyzing requests for the submissions of White Papers for INSCOM (U.S. Army Intelligence & Security Command) Technology when I noticed a number of very interesting requests for a “Content Translation of PAI,” “Automated Sentiment Analysis (SA),” and “Content Generation Based off of PAI”
Let’s give a look at the requirements for each requets:
0001: Content Translation of PAI
A. Capability to translate foreign language content (message text, voice, images, etc.) from the social media environment into English. Required languages are Arabic, French, Pashtu, Farsi, Urdu, Russian, and Korean.
B. Identify specific audiences through reading and understanding of colloquial phrasing, spelling variations, social media brevity codes, and emojis.
1. Automated capability for machine learning of foreign language content with accuracy comparable to Google and Microsoft Bing Translate. Must be able to incrementally improve over time.
2. Recognize language dialect to ensure effectual communication.
0002: Automated Sentiment Analysis (SA)
A. Capability to derive sentiment from all social media content.
1. At minimum, distinguish negative, neutral, and positive sentiment based on collective, contextual understanding of the specific audience.
a) Capability to determine anger, pleasure, sadness, and excitement.
2. Capability to recognize local colloquial and/or slang terms and phrases, spelling variations, social media brevity codes, capitalization, and emojis will be included.
3. Automated machine learning of SA must be able to incrementally improve over time.
a) Software should allow for heuristic updates to improve overall capability; e.g., manually suggest updates based on personal knowledge and experience.
B. Capability to suggest whether specific audiences could be influenced based on derived sentiment.
0003: Content Generation Based off of PAI
A. Capability to translate English into Arabic, French, Pashtu, Farsi, Urdu, Russian, and Korean.
B. Automated capability to generate/create at least three, and up to 10, unique statements derived from one (1) original social media statement, while retaining the meaning and tone of the original.
1. Customize language in a dialect consistent with a specific audience including spelling variations, cultural variations, colloquial phrasing, and social media brevity codes and emojis.
Well, considering that the INSCOM executes mission command of operational intelligence and security forces, this software could be used for both monitoring social media (with a specific focus on Arabic, French, Pashtu, Farsi, Urdu, Russian, and Korean contents).
The requirements specifically address the need to create a system to influence the public sentiment on specific topics (i.e. Capability to determine anger, pleasure, sadness, and excitement.)
I have released this interview to explain my point of view on such kind of software:
What’s the need for such software?
Such kind of software could be used for social media monitoring. Not only, in this specific case, the project aims to build a machine learning system that is able to identify and influence the sentiment of a specific audience. No doubt, it is software that could be used in PSYchological OPerations (PSYOPs).
Would such software be a threat, or would it be just for snooping?
This kind of software could be used also to influence the sentiment on specific topics. This software represents a serious threat to the privacy of citizens, but the most disconcerting aspect is that IA based system could be involved in operations to destabilize a political context. Social media are a powerful instrument to make propaganda and to promote mass demonstration, we have all in mind what has happened in the case of the Arab Spring.
The United States Army Intelligence and Security Command – collects intelligence for combat operations… so how would this new software be useful?
It is an essential software for any Army and Command and Intelligence agency involved in intelligence mission. These systems are essential to gather information from social media and represent a privileged tool for SOCMINT, Social media intelligence.
SOCMINT could be used for several purposes, not only cyber-espionage but also for counterterrorism operations.
Clearly, the most disconcerting use is related to the influence of public sentiment on specific topics.
The U.S. has continually accused Russia of using bots and troll factories to influence U.S. social media… isn’t this an attempt to do the same. What’s your take on that?
I believe that almost any government is involved in social media monitoring, many states use social media as a military weapon. The answer is that the risk is concrete, but at the moment we have no evidence of this. The software could be used also as a defense mechanism against PSYOPs conducted by foreign goverments through social media.
(Security Affairs – US Army, PSYOPS)