11 May, 2020

The efficacy of deepfakes

Can we really write it off as "not a threat"?

   A few days back, NPR put out an article discussing why deepfakes aren't
   all that powerful in spreading disinformation. [1]Link to article.

   According to the article:

     "We've already passed the stage at which they would have been most
     effective," said Keir Giles, a Russia specialist with the Conflict
     Studies Research Centre in the United Kingdom. "They're the dog that
     never barked."

   I agree. This might be the case when it comes to Russian influence.
   There are simpler, more cost-effective ways to conduct [2]active
   measures, like memes. Besides, America already has the infrastructure
   in place to combat influence ops, and have been doing so for a while
   now.

   However, there are certain demographics whose governments may not have
   the capability to identify and perform damage control when a
   disinformation campaign hits, let alone deepfakes. An example of this
   demographic: India.

the Indian landscape

   The disinformation problem in India is way more sophisticated, and
   harder to combat than in the West. There are a couple of reasons for
   this:
     * The infrastructure for fake news already exists: WhatsApp
     * Fact checking media in 22 different languages is non-trivial

   India has had a long-standing problem with misinformation. The 2019
   elections, the recent CAA controversy and even more recently -- the
   coronavirus. In some cases, it has even lead to [3]mob violence.

   All of this shows that the populace is easily influenced, and deepfakes
   are only going to simplify this. What's worse is explaining to a rural
   crowd that something like a deepfake can exist -- comprehension and
   adoption of technology has always been slow in India, and can be
   attributed to socio-economic factors.

   There also exists a majority of the population that's already been
   influenced to a certain degree: the right wing. A deepfake of a Muslim
   leader trashing Hinduism will be eaten up instantly. They are inclined
   to believe it is true, by virtue of prior influence and given the
   present circumstances.

countering deepfakes

   The thing about deepfakes is the tech to spot them already exists. In
   fact, some can even be eyeballed. Deepfake imagery tends to have weird
   artifacting, which can be noticed upon closer inspection. Deepfake
   videos, of people specifically, blink / move weirdly. The problem at
   hand, however, is the general public cannot be expected to notice these
   at a quick glance, and the task of proving a fake is left to
   researchers and fact checkers.

   Further, India does not have the infrastructure to combat deepfakes at
   scale. By the time a research group / think tank catches wind of it,
   the damage is likely already done. Besides, disseminating contradictory
   information, i.e. "this video is fake", is also a task of its own.
   Public opinion has already been swayed, and the brain dislikes
   contradictions.

why haven't we seen it yet?

   Creating a deepfake isn't trivial. Rather, creating a convincing one
   isn't. I would also assume that most political propaganda outlets are
   just large social media operations. They lack the technical prowess and
   / or the funding to produce a deepfake. This doesn't mean they can't
   ever.

   It goes without saying, but this post isn't specific to India. I'd say
   other countries with a similar socio-economic status are in a similar
   predicament. Don't write off deepfakes as a non-issue just because
   America did.

References

   1. https://www.npr.org/2020/05/07/851689645/why-fake-video-audio-may-not-be-as-powerful-in-spreading-disinformation-as-feare
   2. https://en.wikipedia.org/wiki/Active_measures
   3. https://www.npr.org/2018/07/18/629731693/fake-news-turns-deadly-in-india