Monitoring Dissent: How the Jacksonville Sheriff's Office spied on protesters

The Jacksonville Sheriff’s Office used social media surveillance software to track protests against former state attorney Angela Corey, unrest over the fatal police shooting of Vernell Bing Jr. and Black Lives Matter demonstrations until at least late 2016.

The Sheriff’s Office has since cancelled its contract with Geofeedia, which scanned platforms like Facebook, Twitter and Instagram. But the agency stopped short of saying it has halted its practice of spying on civilian social media accounts.

The agency first disclosed its surveillance in 2015, citing police shootings that had garnered national attention. Analysts monitored public dissent with remarkable specificity — down to street corners that were hot spots for protesters.

There are legitimate uses for social media surveillance tools by law enforcement, according to some experts, and the programs are capable of yielding material that can play a key role in criminal investigations and offer evidence in case protests turn violent. But documents obtained by the Times-Union revealed that the Sheriff’s Office also used the software to set up email alerts about First Amendment-protected activity by residents not accused of any crimes. That raises concerns for free speech and privacy advocates over the scope of the Sheriff’s Office’s surveillance network and the potential for abuse of those technologies.

“It’s like they designed their alerts to give civil libertarians a heart attack,” said Rachel Levinson-Waldman, senior counsel at the Brennan Center’s Liberty and National Security program, after reviewing the public records obtained by the Times-Union.

Levinson-Waldman was referring to the broadness of so-called triggers used with the program — words that set off a Geofeedia alert. Social media posts that mentioned “police” or “rally” and were uploaded within designated geographical areas triggered email notifications that went directly to the sheriff’s Crime Analysis Unit.

The Sheriff’s Office declined to detail any of its policies that would restrict or guide the use of social media surveillance software, instead instructing the Times-Union to file a public records request for that information. That request was made Wednesday morning and remained unfulfilled on Friday.

The office wrote in a 2015 report that the monitoring of social media networks had “transcended into focus as the public attitude has shifted toward law enforcement, particularly as it relates to police-involved shootings.” The report indicates that police went beyond monitoring protests and also tracked postings relating to visits by presidential candidates and other potentially contentious events.

Public records further revealed monitoring for chatter among members of a local militia group and mentions of illicit drugs posted from high school campuses.

Police interest in monitoring dissent has persisted. The Sheriff’s Office videotaped a rally against President Donald Trump on the weekend after the November election. The footage was recorded from a distance, but repeatedly zooms in on the faces of speakers taking turns with a megaphone.

‘The cost/benefit was not satisfactory’

A two-day sample of Geofeedia email alerts highlighted that intelligence gathered by the software is not always useful. The alerts were triggered by spam and other postings unrelated to the activity the Sheriff’s Office was trying to ensnare.

An alert meant to monitor a militia group called the Patriot Army returned Jeep auction announcements. Another alert, meant to monitor mentions of the word “rally” on the corner of 9th and Liberty streets, where Bing was shot by police, returned posts about rally car racing. A third, scouring social networks for mentions of the word “bomb,” returned posts about soap, body lotion and fragrant “bath bombs.”

But the alerts were also activated by political postings that would not typically be subject to scrutiny by law enforcement.

“#BLACKLIVESMATTER so #educating my son & #dedicating my energy #time and #knowledge to his survival is Imperative to me,” one captured Instagram post read.

It was collected because of the “blacklivesmatter” hashtag.

“You couldn’t have designed a better case study on how social media surveillance seems to sweep up protected activity on the one hand and perhaps be a significant waste of time on the other,” Levinson-Waldman said.

The Sheriff’s Office disclosed that it had cancelled its Geofeedia contract after declining to fulfill a public records request for all trigger words and alerts set up under the software.

“Its use was discontinued by this agency when it was determined that the cost/benefit was not satisfactory,” Lauri-Ellen Smith, senior spokeswoman for the Sheriff’s Office, said in a statement.

The agency said in October that a yearly payment for Geofeedia totaled $35,000, but that the purchase order for the current fiscal year had not yet been paid. It recently declined to elaborate on what it gained from using the software.

“Our purchases of equipment and hardware/software are transparent to taxpayers, however, the tactical use of that equipment is operational in nature and is not shared,” Smith said.

The Sheriff’s Office said it has no additional social media monitoring contracts aside from its cancelled agreement with Geofeedia, but refused to say it had stopped the practice.

“We will not discuss what we do, tactically, in the arena of social media monitoring,” Smith said.

The Sheriff’s Office said that a Times-Union request for licenses pertaining to all software operated by the Crime Analysis Unit was exempt from public records law.

‘Firehoses’ of data and gigabytes of protest footage

The cancellation of the Sheriff’s Office’s Geofeedia contract came just days after several social media networks — including Facebook, Instagram and Twitter — pulled their data from program.

The company lost access after the American Civil Liberties Union of California in September revealed internal documents from Geofeedia marketing its ability to help law enforcement track 2o14 protests against the police shooting of Michael Brown in Ferguson, Mo., and unrest in Baltimore, Md. over the death of Freddie Gray in police custody the following year.

Although the major platforms have severed ties with Geofeedia, there are myriad monitoring companies that have retained access to “firehoses” of social media data, according to Kalev Leetaru, a senior George Washington University fellow who studies how data changes societies.

Law enforcement agencies could legitimately be interested in monitoring social media around protests to collect evidence in case of a potential violence, Leetaru said. He said recent high-profile protests that have turned violent were instigated by “out-of-town agitators” who could potentially be identified using algorithms.

That offers one possible example of how police could lawfully benefit from collecting data on protesters.

“The trick is, like any tool, it gets abused,” Leetaru said. “There are definitely abuses of these technologies in really bad ways, so that’s what people fixate on and forget there are law enforcement agencies using these tools in a good way.”

‘A great training tool’

While the Sheriff’s Office would not identify its reasons for monitoring protests on social media, it said that it videotapes certain protests and uses the footage as a learning opportunity.

“They are a great training tool to assess how we are responding to events, tactically,” Smith, the Sheriff’s Office spokeswoman, said.

The Times-Union obtained video footage recorded by the Sheriff’s Office of protests held against President Donald Trump on the weekend after the November election that offers anecdotal support for that explanation.

The officer wielding the camera identifies potential pro-Trump “agitators” and zooms out to show the crowd and its movements. The officer also consistently zooms in on the faces of new speakers who take the megaphone.

At least one other major metropolitan police department has used video footage of protesters in conjunction with Geofeedia and facial recognition systems to identify protesters.

Clare Garvie, an associate at Georgetown Law’s Center on Privacy and Technology, said that police in Baltimore used both facial recognition software and Geofeedia, after protests over the death in police custody of 25-year-old Freddie Gray. Garvie said police tracked photos and videos posted in protest areas and ran them through the biometric face-matching database, then sent identities to officers back on the ground.

“The way they pitched it is that they were looking for people with outstanding warrants,” Garvie said.

The Jacksonville Sheriff’s Office has access to facial recognition technology through the Pinellas County Sheriff’s Office’s Face Analysis Comparison Examination System, or FACES. The database includes everyone who has a Florida driver’s license.

Jacksonville police would not identify a policy guiding or restricting its use of facial recognition software when asked by the Times-Union in November. The Center on Privacy and Technology reported that searches of the FACES system are not audited for potential misuse last year.

Similarly, the Sheriff’s Office could not identify any policies restricting its use of social media surveillance software. Though it could not identify written policies, the Sheriff’s Office did say that it does not use facial recognition technology to identify activists using either Geofeedia or protest footage.

There are no state laws, however, preventing police from doing just that.

Research by Garvie and others at the Georgetown center led Republican lawmakers on the House Oversight committee on Wednesday to rail against the privacy risks posed by unregulated use of facial recognition software by law enforcement agencies.

Constitutional concerns

Earlier this month, Facebook announced that it had changed its policy to prohibit developers from using its data to develop tools for police surveillance. But the language has left room for interpretation.

“It’s a great and important step,” Levinson-Waldman said, “But it begs the question of what surveillance means, and how broadly or narrowly that term will be interpreted.”

Social media surveillance of protesters has the potential to violate their First Amendment rights if it tangibly chills them from continuing to participate in that activity, Levinson-Waldman said. She pointed to the Third Circuit Court of Appeals’ October 2015 opinion in Hassan v. the City of New York, which held that the New York Police Department’s monitoring of Muslim communities in New York and New Jersey may have violated the First Amendment in that it inhibited people from continuing to attend their mosques.

As for videotaping protests, Levinson-Waldman said a major concern is the ability to use the video to identify protesters, either visually or using facial recognition software.

“If the protesters are engaging in lawful, peaceful activity and assembly, there is frankly no reason for the government to be in a position to go back and identify who those folks are,” she said.

The Fourth Amendment could come into play, Levinson-Waldman said, in an argument based on how the practice of sweeping up and analyzing troves of data about peoples’ locations, associations, and future plans violates privacy in the same way location-tracing devices do.

“But the law is really mixed, even on the location tracking stuff,” Levinson-Waldman said. “Nobody’s made the argument in the social media realm.”

Ben Conarck: (904)-359-4103