...including this post
1. Recognize a distinction; differentiate.
2. Perceive or constitute the difference in or between.
Discrimination is not illegal or even morally wrong per se, it is an essential part of risk management, but (rightly) there are grounds on which is is illegal to dscriminate in many countries, e.g. race, gender, disability, etc.
For example, my insurance company is perfectly entitled to discriminate by charging me more or less based on what type of car I drive, my age, my driving experience and whether or not I keep crashing in to other people, etc. They are not allowed to charge me differently because of my race or my gender (as recently set down by the EU).
The implication seems to be that Facebook is discriminating (illegally) on the grounds of race, when the available evidence indicates that they are discriminating (legally) on the grounds of geography due to a higher risk of fraudulent access to accounts from some locations. The reasons for higher rates of fraud in certain locations are varied and complex, but poverty juxtaposed with access to global communications have a lot to do with it (419 type scams did not start with email, they started with text messages as mobile phones started to penetrate poorer regions of the world, and actually fax messages were being used in scams even before that).
To coin a phrase, "On the Internet, nobody knows you are a dog" (Google it if you are not of a certain age). All FB has to go on is the login details being used, the IP address of the machine (and therefore a way to infer your geographical location) and a fair bit of technical information about your browser plus any evil zombie tracking data the Overlords have managed to secret on the machine. Put simply, if a non-Kenyan went to Kenya and logged on to FB from a local machine, would they face the same security challenges? The evidence presented indicates that they would.
I suppose it might be possible that FB based its geographical risk profiling on illegally discriminatory criteria, i.e. "Most people in country X are of race Y and I therefore more likely to be criminals in my view, so I'm going to increase the risk rating in the database for IPs in that location", but nothing in this article provides any evidence that this is the case.
Knowing how naiive many of these Web 2.0 businesses are in so many areas, it wouldn't be a surprise if the basis of their risk profiling *was* down to the predjudices of some wet-behind-the-ears college grad rather than empirical (and legally defensible) data. Now that would make an interesting story, but this article does not.