When it comes to implementing policies to address that, you look at who the systemic problems are affecting and you make choices to help address those effects. I'm sure some people with some arbitrary SSN are a person who is affected by institutional oppressions, but people are not targeted for inclusions in those systems based on SSN, so that's not a metric worth targeting when implementing policies (although if it can be demonstrated that there is a significant link between SSN and institutional racism, that's another story).
There are lots of other external properties that people are generally prejudiced for or against, such as weight/height/build, (dis)ability, posture, voice/speech properties, dress sense, and so on; but these (a) are considered to be more under an individual's control, (b) aren't inherited, and (c) historically haven't caused anywhere near as many social problems as racism in the US. No doubt people who are discriminated against based on their voice (say) don't like it, but it's not considered to be a systematic, self-reinforcing, widely-observed, entrenched social problem.
A hypothetical prejudice against "SSN % 104 == 7", where the property is not even outwardly observable (so can't genreally taint initial impressions), nor subject to this ongoing reinforcement, nor passed down through generations (neither the prejudiced property, nor the prejudice itself), is completely different from race, even moreso than the other examples.
Statistically, group A has been and is on the receiving end of many institutional oppressions that keep them excluded from accessing higher education. We can tweak the definition of A a bit if it makes the analogy easier:
Group A = people who are either victims of racism directly OR people with SSN % 104 == 7.
Some members of group A have suffered racism, others have not. Just like group B. Both groups are statistically more likely to be racism victims, but plenty of people in those groups have not suffered racism in any significant way.
How do you distinguish between group A and group B? None of the criteria you have stated here actually differentiate between them.