Trending: Why net neutrality’s peril raises the stakes for future satellite broadband options

Manka Dhingra is a Democrat running for Washington state Senate. (Photo via the Dhingra campaign)

Update: Manka Dhingra was ahead with 55 percent of the vote in early election results released Tuesday at 8 p.m.

Bots and other suspicious social media accounts were used in an apparent attempt to influence voters in a heavily funded Washington state Senate race that could tip the balance of political power in the Western United States, according to a report from a non-partisan group of experienced cybersecurity researchers.

The report, which the researchers provided to GeekWire today, cites extensive evidence that suspicious Twitter accounts have been advancing attacks against Manka Dhingra, a Democrat running against Republican Jinyoung Englund in Washington’s 45th District, which includes parts of Kirkland, Redmond and Sammamish as well as other cities in the suburbs east of Seattle. After GeekWire reported on the suspicious Twitter activity, a former volunteer for Dhingra alerted us to fake accounts on Facebook as well.

Jinyoung Englund is the Republican candidate in Washington state’s 45th District.

The researchers behind the report, whose identities and credentials are known to GeekWire, declined to be identified publicly, saying that their broader investigation into “computational propaganda” in the U.S. and around the world would be compromised as a result.

The Election Day revelation is the latest twist in a race that has become a battleground for outside interests. A victory by Dhingra would flip Washington state’s Legislature and solidify all three West Coast states into a Democratic stronghold, including the governor’s offices and legislatures. The stakes have put an ordinary state race into the national spotlight, drawing more than $9 million in campaign contributions, the most expensive legislative race in Washington state history.

It’s a sign that online propaganda, a cloud hanging over elections nationally and worldwide, is also being wielded in an attempt to influence the outcome of state and local races.

In their report about the 45th District, the researchers say that almost 10 percent of accounts tweeting about Englund and Dhingra over the past 30 days were classified as bots, accounts that use automation to spread messages on Twitter. “There is evidence of coordinated activity among the suspicious accounts tweeting aggressively against Manka,” the researchers write.

There’s no evidence of any illegal activity, in part because lawmakers haven’t caught up to the new ways social media is being used to spread disinformation.

The report does not identify any people or groups suspected of orchestrating the social media activity in the pivotal Washington state Senate race. However, the researchers point to the broader influence of corporate donors on the 45th District campaign.

“Funders on both sides have run divisive ads, but the fear-mongering tone of advertisements against Dhingra and exploitation of wedge issues from right-wing funders have been particularly overt this cycle,” they write in the report. “Englund’s campaign has also taken money and received support from organizations that have a documented track record of divisive and sometimes misleading campaign tactics, from big oil and big tobacco to the Koch Industries.”

With this backdrop, the report investigates whether some of this outside money “is being used for computational propaganda supporting Jinyoung Englund and deriding Manka Dhingra.”

GeekWire has contacted the campaigns for both Dhingra and Englund for comment, and we will update this story with their responses.

According to the researchers, the suspicious accounts were identified by pulling a sample of 850 accounts that tweeted about the race during the past 30 days. Researchers ran the accounts through two different open-source algorithms and then manually examined them, classifying 22 as likely to be using automation. The accounts were primarily pushing content aligned with pro-Englund campaigns.

Editor’s note: This post was updated to clarify how the suspicious accounts were identified. 

One researcher told GeekWire that this is “the tip of the iceberg” because the findings were limited to publicly available data on just one social media platform.

Update: GeekWire has learned that suspicious online activity targeting Dhingra was not limited to Twitter. Nicholas Duchastel de Montrouge, a software engineer for Dropbox who volunteered for Dhingra, said he and others in his network identified several fake Facebook accounts impersonating the candidate. He also spotted several sponsored posts from the fake accounts that spread critical or false information about Dhingra, like the one below.

Duchastel de Montrouge filed a complaint with the Washington State Public Disclosure Commission and was told there wasn’t much the organization could do. He says that some of the impersonator pages have been taken down. The name on one page that is still up has been changed from “Manka Dhingra” to “Unofficial Manka Dhingra” and added the legal disclaimer “paid for by Working Families,” Duchastel de Montrouge says. Working Families is a Political Action Committee funded by The Leadership Council, an organization that has received money from Republican groups.

Before Dropbox, Duchastel de Montrouge was an engineer for Facebook’s advertisement team in Seattle. He’s frustrated that his former employer isn’t doing more to crack down on imposter accounts.

“I know for a fact that any engineer at Facebook could figure out who created these fake pages, fake posts, and paid ads,” Duchastel de Montrouge said in an email. “This takes less than 10 minutes.”

Below are screenshots of posts from accounts posing as Dhingra, provided to GeekWire by Duchastel de Montrouge.

“It’s important to remember that the most sophisticated computational propaganda techniques don’t rely entirely on bots, but combine human and machine actors in a multi-modal, multi-channel strategy to distort public conversation,” the report says. “In the context of the 45th race in Washington State, our research found that digital propaganda is multi-modal in nature: we see similar behavior from known bot accounts, trolls, and paid operatives.”

Researchers used machine learning algorithms to study a wide range of factors, such as frequency of posts sent, post content, images, and others, which taken together could signal an automated account. If for example, an account says it is in Washington but frequently tweets at 4 a.m. PST, that’s a red flag.

Suspicious accounts have frequently attacked Dhingra’s support for safe injection sites over the past month, according to the report. That’s a narrative that Political Action Committees (PACs) supporting Englund have also been using. Researchers found evidence that automated accounts have been driving traffic to “slanted and propagandistic ads” against Dhingra. Several of the accounts identified link to SafeKingCounty.org, an anti-safe injection site paid for by IMPAC Washington.

The tweet below comes from an account researchers say has a “strong probability” of being a bot.

The online efforts to paint Dhingra as an extremist signify that digital propaganda campaigns have spread from high-profile national elections to local races. This has become a hot-button political issue over the past year, as evidence surfaced that Russian agents used social media to influence the 2016 presidential election. The role technology played in electioneering was the subject of a series of tense Congressional hearings where lawmakers grilled lawyers from Facebook, Twitter, and Google earlier this month.

“This is something that’s going on on a statewide level now,” said one author of the report. “It’s not just national or international, it’s also localized in a way that’s really interesting.”

One of the researchers said the group has watched online electioneering trickle down from big national races to local ones and they don’t anticipate the trend to slow without new regulations to curb social media propaganda.

That fits the narrative that Duchastel de Montrouge heard when he tried to report suspicious activity and imposter accounts to Facebook.

“I reached out multiple times through the website, I reached out through two to three friends that still work at Facebook, and it’s always the same thing,” he said. “Unless you give us a court document, you can’t do anything.”

Like what you're reading? Subscribe to GeekWire's free newsletters to catch every headline

Comments

Job Listings on GeekWork

Data Engineer (crowdsourced, linguistic)The Allen Institute for Artificial Intelligence (AI2)
Find more jobs on GeekWork. Employers, post a job here.