“Trust but verify” – Russian proverb.
Earlier this year The Attributer published an article entitled ‘Fake Protected’. This new article revisits the same topic but from a different perspective. We are living in a time of increasing outbreaks of ‘fake news’ in particular, and ‘fake information’ in general. How do we establish the provenance of data and the information that it encodes? Perhaps the more important question to ask is: are we looking at the wrong metadata? The issue is exploitation of trust without verification. This problem undercuts our fundamental security foundations.
One of the earliest trust-with-verification mechanisms was the handshake. Since most of the world is right-handed, the extended open palm of the right hand signalled that both parties were unarmed, therefore the handshake was safe. In the virtual environment where only some senses are invoked, the ability to verify is compromised. There is no eye contact, no body language and no assessment of dress codes and other signals of personal identity and integrity. This phenomenon is witnessed frequently with autonomous systems where subtle changes to signs and other environmental objects become significant problems for systems that cannot verify content.
Content can be verified when contextually examined. This activity requires not only more effort, but also re-thinking our security paradigm and designing new solutions that capture contextual information. Capturing contextual data provides the ability to gain new insights from old events.
Consider the case of fake news. There are three critical characteristics of ‘news’ that can be examined: computational linguistics (CL), pattern-spread (PS) and source provenance (SP). CL can be used to identify and quantify linguistic markers of deceptive speech. PS can use signal gathering techniques to identify normal and anomalous propagation patterns, and SP can be used to examine the history of the author and publisher.
The Oxford English Dictionary defines provenance as: ‘the place of origin or earliest known history of something’. As a secondary definition it says: ‘the beginning of something’s existence; something’s origin’. In one of the sentence examples it gives this usage: ‘False provenances and certificates of authenticity are favourite tools of cheats and should never be accepted blindly.’ You bet!
In 2017 a report on fake news mentioned that discrediting a reporter can be done for $50,000 and a fake protest can be created for $200,000. The relatively low cost associated with both of these acts suggests that they will become more frequent. Serious journalists may be fooled once or even twice, but not repeatedly. These same journalists will have a body of work that can be examined through archived data by using CL to classify their work. The work history can be scored and trended indicating whether the veracity score of the reporter’s work is improving or worsening.
A Jigsaw Research (2018) found that while the majority of people rely on the television, followed by the Internet as sources for news, a full 82% of young people (16-24 years) prefer the Internet. The same Internet that is vulnerable to injection of non-verified data as discussed above. Furthermore, when trusted sources are attacked and discredited the trust model further erodes. This results in sentiments listed below:
“The only people I trust to tell me the truth are my Facebook friends”.
“I only use Facebook and Twitter to get news and information about the world. I don’t trust the other media”.
The truth about fake news is complex. Consider the BREXIT referendum where voters, young and old, educated and uneducated, supported leaving. The messages were crafted to appeal to the voters’ deeply held values and beliefs. These messages were crafted to demographic groups identified by the very social media that people increasingly rely on for their news.
Countering fake news will require reliance on technical and non-technical solutions. The ability to critically assess information in this digital world is a topic that should be included in every school and college curriculum. The issue is to do with three major things: what is the source of the information? Do you trust that source? Can you verify the data? It will require more SABSA Thinking.
(With thanks to Dr. Char Sample for her advice)