Nafeez Ahmed, Motherboard:
The United States government is accelerating efforts to monitor social media to preempt major anti-government protests in the US, according to scientific research, official government documents, and patent filings reviewed by Motherboard. The social media posts of American citizens who don’t like President Donald Trump are the focus of the latest US military-funded research. The research, funded by the US Army and co-authored by a researcher based at the West Point Military Academy, is part of a wider effort by the Trump administration to consolidate the US military’s role and influence on domestic intelligence.
It’s no secret that the Pentagon has funded Big Data research into how social media surveillance can help predict large-scale population behaviours, specifically the outbreak of conflict, terrorism, and civil unrest.
Much of this research focuses on foreign theatres like the Middle East and North Africa — where the 2011 Arab Spring kicked off an arc of protest that swept across the region and toppled governments.
Since then, the Pentagon has spent millions of dollars finding patterns in posts across platforms like Facebook, Twitter, Instagram, Tumblr, and beyond to enable the prediction of major events.
But the Pentagon isn’t just interested in anticipating surprises abroad. The research also appears to be intended for use in the US homeland.
In August, a US Army-backed study on civil unrest within the US homeland was published in an obscure anthology of papers presented to a Big Data conference in Kiev, Ukraine, which took place in early June. The anthology was released as part of Springer-Nature’s Advances in Intelligent Systems and Computing book series.
Datasets for the research were collected using the Apollo Social Sensing Tool, a real-time event tracking software that collects and analyses millions of social media posts.
The tool was originally developed under the Obama administration back in 2011 by the US Army Research Laboratory and US Defense Threat Reduction Agency, in partnership with Rensselaer Polytechnic Institute, the University of Illinois, IBM, and Caterva (a social marketing company that in 2013 was folded into a subsidiary of giant US government IT contractor, CSC).
But the use of the Apollo tool to focus on protests in the US homeland has occurred under the Trump administration. The ‘election’ dataset compiled using Apollo for the 2018 US Army-funded study is comprised of 2.5 million tweets sent between October 26, 2016, and December 20, 2016, using the words “Trump”, “Clinton,” and “election.”
Tweets were geolocated to focus on “locations where protests occurred following the election” based on user profiles. Locations were then triangulated against protest data from “online news outlets across the country.”
The millions of tweets were used to make sense of the “frequencies of the protests in 39 cities” using 18 different ways of measuring the “size, structure and geography” of a network, along with two ways of measuring how that network leads a social group to become “mobilized,” or take action.
The paper concludes that by “examining the structure of social networks as related in tweets related to the 2016 US Presidential Election, a relationship is identified between network structure and protest occurrence.” The model demonstrates that social media plays a catalyzing role in the mobilization of social groups before a protest, in a way “which is observable in advance of protest occurrences.”
More work is needed to beef up the accuracy of the model, though. This model is still only 44 percent accurate five days before the protest, with accuracy increasing up to 82 percent closer to the incident.
Whatever the case, the US military funding remains a clear indication of US government interest.
The Rensselaer Polytechnic Institute is part of the ARL Network Science Collaborative Technology Alliance (NS CTA), a consortium of three industrial research labs and 14 universities which receives multi-million dollar support from the US Army Research Laboratory.
The last round of multi-million dollar five-year funding was received by Rensselaer in 2015. Research priorities are closely and continuously developed within the CTA through collaboration between university scientists, industry and the US military.
A lead author of the paper, Rostyslav Korolov—identified as the point of contact about the research—is a PhD candidate at RPI focusing on “prediction of human behavior based on social media communications” under the ARL Alliance, and liaised closely with the US military while working on the anti-Trump protest study.
“While working on this project I’ve spent two months on an internship at the US Army Research Laboratory and a year as a visiting scholar at the Network Science Center, United States Military Academy, West Point,” his RPI bio explains.
Motherboard attempted to reach the authors of the paper through multiple requests, but did not receive any response to questions about the study and the reason for its focus on protests at home.
Price is the world’s leading expert on the relationships between US anthropologists, social scientists, and US military intelligence agencies, the author of the book Weaponizing Anthropology: Social Science in Service of the National Security State, and has served on several American Anthropological Associations commissions and task forces dealing with the ethical issues of engaging with the US intelligence community. He describes the latest research as “really an extension of the once frightening, now mundane expression of a national panopticon expressed by the Total Information Awareness program when first conceived in 2002, and quickly withdrawn.”
Total Information Awareness was a major Bush administration initiative aimed at monitoring the entire American population through electronic surveillance. Though defunded in 2003 after extensive media criticism, its core architecture was adopted by the National Security Agency (NSA) where it is now “quietly thriving” according to the New York Times.
Established in April 2012, the project (which stands for Early Model Based Event Recognition using Surrogates) generated seven-day advanced forecasts based on “open-source indicators”—social media, satellite imagery, and more than 200,000 blogs that are publicly available. An average of 80 to 90 percent of its forecasts were accurate, according to studies related to the program.
Teams made up mainly from three external industry partners, HRL Laboratories, Raytheon BBM Technologies, and Virginia Tech, were involved in developing the technologies behind Embers, which was funded by a $22 million contract by IARPA.
In short, in 2017 the Trump administration moved IARPA’s Embers social media surveillance program into the private sector under VTARC. Yet one of VTARC’s customers using these surveillance tools is the Trump administration, and based on the job listing, it appears to deal with secret and top secret information.
The move by VTARC illustrates that even with the best of intentions, independent scientists receiving US government funding for such research have no control or oversight over the uses of their work. According to Price, the impact of the research could still be insidious even if the social scientists involved did not hold any conflicts of interest as such.
“This sort of military funded social science research tends to occur in an ideological echo chamber, where groupthink predominates and dissent or concerns about the applications of this work is missing,“ he told me. “Among the basic assumptions that social scientists outside this group would question are assumptions that civil unrest or protests are not core elements of democracy that need to be protected, [rather than] undermined by surveillance—and the oppression that follows such surveillance.”
VTARC did not respond to requests for details on who the corporation’s clients are for its social media surveillance tools originally developed under Embers.
Earlier this year, the ACLU filed several FOIA requests to a range of US government agencies over concerns that domestic social media surveillance had “spiked” under the Trump administration. In early September, the ACLU released documents showing that state and federal law enforcement agencies were collaborating to ramp up “Social Networking” surveillance of domestic activists over concerns about protests against the Keystone XL pipeline—the measures were justified as “anti-terrorism.” Two weeks later, an image contained in a Massachusetts state police tweet accidentally revealed that a police computer monitor had bookmarked several Facebook groups for left-wing activist organizations with an anti-Trump slant.
While the ACLU has been able to confirm that under Trump, government departments like the Departments of Defense, Justice, and Homeland Security are accelerating domestic social media surveillance in relation to anticipated anti-Trump protest incidents, these FOIA requests have not revealed the technologies being deployed to do so.
HRL is jointly owned by General Motors and Boeing. The successful patents relate to a whole ecosystem of social media surveillance technologies, many of them still in application, developed over nearly a decade with funding from IARPA.
One patent is titled “Tracking and prediction of societal event trends using amplified signals extracted from social media,” filed in 2013 and granted in February 2018. The invention, says the patent, relates to “a system for tracking and prediction of social events using amplified signals extracted from social media.”
Another patent is titled “Inferring the location of users in online social media platforms using social network analysis,” filed in 2013 and partially granted in October 2017.
The body of scientific literature related to these patents, reviewed by Motherboard, demonstrates a sophisticated technology suite capable of locating the “home” position of users to within 10 kilometers for millions of Twitter accounts, and predicting thousands of incidents of civil unrest from micro-blogging streams on Tumblr.
Although these technologies were developed under the Obama administration, it appears their use is being accelerated by the Trump administration—and by moving the Embers program to which these technologies relate into the private sector, this acceleration is occurring in a way that sits beyond public scrutiny or accountability.
The intensification of US social media surveillance coincides with the Trump administration’s augmentation this year of the Pentagon’s role in homeland security.
In April 2018, the US Joint Chiefs of Staff issued an updated doctrine on homeland defense. The new doctrine underscores the extent to which the Trump administration wants to consolidate homeland defense and security under the ultimate purview of the Pentagon.
The US military’s traditional function is to defend the US from foreign threats rather than interfere with domestic issues. After 9/11, homeland defense and security doctrines have been gradually pushed toward closer integration with the US military in a process first accelerated by the Bush administration.
As with previous versions of the doctrine, the document states that ‘Lead Federal Authority’ (LFA) for “homeland security” is the Department of Homeland Security, but simultaneously goes to pains to emphasize again and again how the Department of Defense (DOD) must be active at the epicentre of almost all homeland affairs.
These stipulations are not entirely novel compared to previous iterations. Yet they are augmented by some subtle but unprecedented changes concerning the powers to respond to a domestic “insurrection” and the role of Pentagon intelligence in such a response.
Crucially, Aftergood pointed out that some of the most notable changes in the doctrine concern ensuring that classification does not prevent homeland agencies from accessing Pentagon intelligence. The upgraded doctrine says that Pentagon resources can be mobilized for domestic surveillance or “information support” in the context of emergencies.
“Military information support forces and equipment may also be used to conduct civil authority information support activities during domestic emergencies within the boundaries of the US homeland,” it reads.
In simpler terms, the doctrine insists that classification should not impede the Pentagon from sharing intelligence with domestic agencies, especially in the context of “civil authority information support“ in homeland emergencies.
According to Price, this has ominous implications given that the NSA is a DOD agency. The import is that under the new doctrine, there are greater opportunities to connect domestic intelligence gathered by the NSA with the social media data of American citizens.
The most pertinent section of the upgraded homeland defense doctrine for this story concerns the powers available to the President in the case of “insurrection,” a major rebellion which either state governors or the President himself deem to fundamentally threaten the rule of law.
Once again, though not a new addition to the doctrine, vaguer language is introduced including a stipulation that the authority for a US Army response to an “insurrection” will constitute a “HD [homeland defense]-related purpose.”
This is the first time that an “insurrection” has been described using the phrase “homeland defense,” implying that the response would come under Pentagon jurisdiction.
I asked Banks, co-author of Soldiers on the Homefront: The Domestic Role of the American Military, about the doctrine’s description of an “insurrection” as a “homeland defense“ issue.
“The US military role in the homeland is not new, but in this case there’s a tension between DSCA [Defense Support for Civil Authorities] and homeland defense, because in one setting civilians are in charge, and in another setting the military are in charge,” he said. “The changes to doctrine are not dramatic, but they could make it more likely, maybe inevitable, that those jurisdictional issues might come together or clash in some way.”
The outcome of such a clash could end up putting Trump’s Defense Secretary in charge of a response to a domestic emergency categorized by Trump as an “insurrection.“ Taken in tandem with the US military’s sudden interest in predicting anti-Trump protests after the 2016 elections, the Pentagon’s upgraded homeland defense doctrine seems to be part of a wider effort by the Trump administration to prepare for domestic civil unrest in coming months and years.
The Pentagon did not respond to Motherboard’s question about any possible connection between the upgraded homeland defense doctrine and the Pentagon’s new research on social media surveillance.