US Intelligence Community Assessment of 2020 election interference. Cruder disinformation. COVID-19 misinformation. Deepfake blues.

At a glance.

• Influence operations and the 2020 US elections: Russian operations.
• Influence operations and the 2020 US elections: Iranian operations.
• Influence operations and the 2020 US elections: Chinese operations.
• Influence operations and the 2020 US elections: minor players (and criminals).
• Disinformation campaign hijacks Polish government sites, and a journalist's Twitter account.
• Jamming can be harder than it looks.
• COVID-19 misinformation.
• Deepfake fears, including the AI training supply chain.

The US Intelligence Community this week released an unclassified version of its report on foreign interference in the 2020 Federal elections. The investigation found no evidence of foreign attempts to manipulate vote counts or other “technical aspects” of the election. That said, there were some significant foreign efforts to influence US voters. Here's how the influence horse race turned out.

Election influence operations: a win for Moscow.

The most extensive and arguably the most influential disinformation campaign was mounted by Russia. To call it a "win" may be a bit of an exaggeration, given the outcome of the election, but give the Russian organs style points: their influence campaign was complex, sophisticated, long prepared, and carefully thought through. And given that a major strategic objective was to exacerbate fissures and sow mistrust in American civil society with a view to weakening Russia's biggest adversary, it seems only fair to let Moscow chalk up a W.

The Intelligence Community Assessment found evidence of an extensive Russian influence campaign aimed at denigrating then-candidate Biden to the advantage of then-President Trump, with a strong overarching goal of eroding confidence in US elections. The effort was authorized by President Putin and other senior Russian officials.

Russia’s efforts were marked by extensive preparation and the use of trolls, agents of influence, and American influencers of the useful-idiot variety, with messaging amplified by online proxies and Russian official media outlets. In general Russian policymakers, while not in every respect happy with President Trump, clearly preferred him to a President Biden, although they had made their peace with a possible Biden Presidency by the closing weeks of the campaign, seeing a silver lining in President Biden’s presumed interest in reviving arms control agreements perceived as working to Russia’s advantage. Their longstanding goal, which the report says endures into the present, is to weaken the United States, and whatever is likely to accomplish that, particularly erosion of trust in US civil and political institutions, is a good bet.

As one might expect, the Russian Embassy in Washington didn't much like the IC’s report: "another set of groundless accusations against our country of interfering in American internal political processes." The report, says the Embassy, is just more American "megaphone diplomacy."

Election influence operations: Tehran places.

Iran conducted a similar influence effort aimed at damaging President Trump’s candidacy. Like the Russian disinformation campaign, Iran's efforts were authorized at the highest levels, in this case by Supreme Leader Khamenei.

Iran wasn’t particularly in favor of President Biden, but the Islamic Republic was definitely opposed to President Trump. Their influence operation ran principally through social media and, interestingly enough, through highly targeted email campaigns that spoofed rightwing groups like the Proud Boys and threatened the recipients, for the most part likely Democratic voters, with crude appeals to vote for Trump, evidently hoping thereby to provoke a backlash against the former President. Tehran’s efforts worked to exploit and exacerbate fissures in American civil society, and the report warns that these efforts have continued, post-election. Iran chose what the report calls “cyber tools and methods” because those methods were cheap, scalable, deniable, and required no physical access to the US.

Election influence operations: Beijing shows.

The investigation considered the possibility of interference by other governments as well, but none of the others were as active as those of either Russia or Iran. China is assessed as having considered undertaking an influence campaign, but eventually seems to have decided to sit the 2020 election out, apart from taking some minor shots at then-President Trump.

Beijing seems to have conducted a cost-benefit analysis and decided that it saw no particular advantage to China in the election or defeat of either major party candidate. Thus there was judged to be no significant payoff in either possible electoral result. And the costs, particularly in the form of potential blowback should they be caught attempting to push either candidate across the finish line. Being caught finagling would have been bad optics. Traditional influence--lobbying and economic incentives (or disincentives)--were judged to be the best bet for advancing Chinese interests.

In any case, the view from Beijing sees a bipartisan Sinophobic consensus in the US, and holds that such anti-China sentiment is likely to endure whichever party holds the major positions in Government. Beijing may have thought President Trump mildly worse for Chinese interests than President Biden, but not worse enough to warrant a big push to see him defeated.

Election influence operations: the also-rans.

Lebanese Hizballah, Cuba, and Venezuela played bit parts with their own minor influence operations. None of them had any use for President Trump and woofed against him, but their efforts were ineffectual, petty larceny stuff, lost in the noise. And of course there was the usual criminal presence, manifesting itself in ransomware attacks, at least one of which affected a voter registration system. But the crooks don’t appear to have been aligned with any government, nor to have had any particular political purpose. The crooks just want to make bank. And the hacktivists, to round out the third conventional category of threat actor? They were hard to find, apart from some freelancing Turkish patriots who did some hard-to-notice baying in what they took to be Ankara's interest.

That radiation leak? Never happened.

Not all Russian disinformation is as sophisticated or effectual as that deployed against the US elections. A contrasting case in point emerged in Eastern Europe this week.

Poland’s government has provisionally attributed a disinformation effort about a bogus radiation threat to Russia, the Washington Post reports. There were three channels for the propaganda: websites of the National Atomic Energy Agency and the Health Ministry were compromised to briefly display fabricated claims of nuclear waste leaking into Poland from neighboring Lithuania, and a Twitter account belonging to a journalist whose beat is Russia and Eastern Europe.was also hijacked to push the same story.

It is, of course, bogus--there’s no radiation leak in Lithuania, and there’s no corresponding threat to Poland. The Polish government representative who attributed the incident to Russia did so on grounds of a priori probability, but it’s a pretty good guess as an argument to best explanation. Stanislaw Zaryn, speaking for the head of Poland’s security services, told the Associated Press that “‘the whole story looked like a typical Russian attempt’ to sow suspicion and division among Western allies.”

The operation is notably more heavy handed than the Russian disinformation the US Intelligence Community described. It involved compromising websites, for one thing, and hijacking Twitter accounts for another. Such methods are of only short-term effect, because website defacements and social media account hijacking are fairly obvious, "noisy," and are usually swiftly detected and taken down. They might serve to throw a temporary scare into a target, or as a kind of artillery preparation for a deeper attack against the credibility of otherwise trusted sources.

How hard can it be to shut down objectionable content in Twitter? Harder, apparently, than one might think.

We saw, last week, how Russia had threatened to slow Twitter down unless the platform got better at interdicting content the government found objectionable, criminal. (In fairness to Roskomnadzor, Russia's Internet authority, the nominal targets were things most civilized people would want little part of, like suicide, drug abuse, and child exploitation, although other, political censorship is also widely suspected as an underlying motive.) Rest of World reports that the slowdown affected Internet performance generally, impeding other activity as a kind of collateral damage. Meduza offers some informed speculation about how deep packet inspection may have failed Roskomnadzor this time around. But the authorities are undeterred. The AP says that Moscow has renewed its threat and upped it from a slowdown to a blockage. Twitter has a month to get right with the Kremlin.

COVID mis- and disinformation.

Axios reports that, over the last six months, YouTube removed some 30,000 videos that featured COVID-19 misinformation. Crafting rules for such content moderation is difficult, as Facebook continues to find, and applying such rules consistently is equally challenging. The Washington Post says that Facebook has found that influencers will influence; that is, a relatively small number of individuals have a disproportionate effect on the opinions of those who consume their pensées.

Deepfake blues.

So there's some good news about deepfakes. First, Scientific American reports that an IBM AI, "Project Debator," went up against a human component in a formal debate and lost, so Project Debator isn't, at least for now, as good at debate as Deep Blue got to be at chess. And for now, again, deepfake images still bear some characteristic stigmata that betray them for what they are: the reflection of light in the eyes of the people whose pictures appear in a deepfake look funny. Researchers at the University at Buffalo (the institution formerly known as SUNY Buffalo) found that they could reliably recognize a deepfake by its unconvincing reflections. (It's vaguely troubling that they wrote a program to do it, but for now at any rate, it seems possible to see through the deepfakes.)

But there's bad news, too. Artificial intelligence requires data--a lot of data--for training. IEEE Spectrum points out that this amounts to a supply chain. What effect might deliberately corrupted data have on the performance of otherwise benign AI?