I've had an interesting day today. Checked into the airport this afternoon, and had a debate with the woman at the counter about my reservation. I received my ticket and was surprised to see I wasn't sitting in business class.
The funny thing is, I had an itinerary and record locator that indicated that I was in business class, but our check in clerk claimed I didn't.
A quick call to her supervisor came back with a confirmation that I did not have a business class seat. The options - take a business class seat for another $200 Euros or take a seat in coach. There was some additional discussion on my part, but I was amazed at how uninterested and unhelpful this particular individual was.
Before leaving the desk, I requested that she use my air miles card from a partner airline. Her response, which struck me as a bit odd, was that there was no need, as I was a gold member.
I begrudgingly took the coach seat and made my way to security. While in line I was thinking about her comment about my being a gold member. While I'm gold on other airlines, this (and the partner) weren't one of them.
I rechecked my ticket. Not sure who Vincent Mercier is, but he sounds a bit more French than this guy who grew up in Tewskbury, MA and knows just enough French to be either polite or offensive. I returned to the desk, pointed out the mistake and had my business class ticket in hand.
When sitting in the airport lounge a bit later, I thought about what had just transpired. Air France has asked initially for my passport, to check claims of identity. That claim wasn't used successfully by the requestor, and a secondary claim - my reservation locator - was provided. Again, this wasn't used. Without success, the workflow required an escalation to another service - the supervisor - and again there was a failure. Here it was based on the information provided by the initial requestor.
It stresses the potential for a breakdown in an identity valdation scenario which involves a human component. The difference between Vincent Mercier and Marc Mercuri is fairly obvious, but the check-in cleark may have done some faulty pattern recognition based on seeing MERC in both.
Had this been a machine driven interaction, this would likely have gone flawlessly. A selection of destination city would have been used to limit the number of potential name matches and from that subset, the name would have been valdated either 1:1 or possibly with something along the lines of a Soundex).
What makes this breakdown of 'the system' incredibly alarming is that there was no validation of claims from that point forward - once ticket was in hand, I had free access to the system, boarded the plane, disembarked in Paris and am now in my hotel.
Sure, I provided the token assigned by the airline (a boarding pass) at security - but there was no requirement/check of my passport. If I had continued through with my initial, erroneously issued token (the ticket in someone elses name), I would surely still be in Paris eating the French interpretation of Cajun Chicken wings.
In this specific context, an identity breakdown has horrific potential. Suppose the mistaken identity had occured not with a guy less interested in connecting systems as disrupting them - a terrorist.
There were no further checks for identity (intra-EU flights do not have passport control), so someone who slipped through the system could now be freely traversing France. Given the political climate here in Paris this week (for those unaware, there have been riots and indivduals setting fire to cars in France), it's even more alarming.
With the recent move to self-service kiosks for check in, the mechanisms I mentioned earlier are helping avoid this issue. Introducing some of the technology used there in the human interaction piece (i.e. scanning of passports and system retrieval of information) would help solve the issue, surely.
But that answer begs different questions. We do quality assurance of the software systems, but how do we and how much time do we do testing of the human components in connected systems? And once you've established your test plan, and you go to 'rtm' of the process/workflow, how do your federated users report bugs? In this particular instance we're not talkng about a situation that results in some bizarre behavior in an IDE, we're talking about international security in the heart of Europe. The clerk surely isn't going to tell her manager, as it points out big mis-step on her part. There's no contact information on the boarding pass or airline timetable. Going to the Air France web site, I went to the link to their corporate office, which is entirely in French. I'm on a hotel internet connection at 90 cents per minute, chances are I'm not going to spend an hour navigating their site to let them know about the issue, resulting in an open loophole in a frequently used workflow with potential for failure far, far worse than any blue screen.
In this particular scenario, the issuance of a false token was an 'honest mistake', but suppose that it wasn't. Imagine if a terrorist cell had someone working behind the ticket counter, what checks are in place to prohibit <i>intentional</i> bad issuance or trust violations?
This isn't just with transportation companies, it spans verticals. For example, if John Smith is caught owing $200,000 in taxes, and the workflow for resolving this dispute is handled by Bill Jones who makes $20,000 per year, John paying $50,000 to Bill Jones to make this whole matter disappear happens. Depending on the country, it happens alot.
These example involved a relatively simple workflow, this obviously gets more complex when dealing with interactons that run multiple partners/parties deep.
If you have a business with a high volume of transactions or high value transactions with consumers or areas with complex workflows , how do you / would you handle these situations? What types of SLAs and legal terms do you have in place to handle scenarios where a human taints the system with a manual violation of trust in a federated scenario?Feel free to speak in the third person and without corporate identities, I'm curious how/if this is being addressed.