Major Bug in libpax detected?

[commeco]

Hi there,
Sorry for posting here, even it's more of a LibPax topic. But there's a much higher chance of getting an answer here. :)

We used a prototype of the PaxCounter with an enhanced scan window of six hours for testing purposes.
I was surprised by the counting values (very high), so I spent some time investigating in the LibPax.
I found the following line in the code:
https://github.com/dbinfrago/libpax/blob/3cdc0371c375676a97967547f4065607d4c53fd1/lib/libpax/libpax.cpp#L77

Definition of ("universal" and "local" mac addresses)
"The second least significant bit of the first byte determines whether a MAC address is universally or locally administered."

So...
0 (even) | Universally administered (assigned by vendor)
1 (odd) | Locally administered (likely randomized)

  // if it is NOT a locally administered ("random") mac, we don't count it
  if (!(paddr[0] & 0b10)) return false;

Besides the fact that also the comment is incorrect, this if statement filters all real MAC addresses and returns false. The 'not' operator is the problem.

I would suggest:

  // if it is a locally administered ("random") mac, we don't count it
  if (paddr[0] & 0b10) return false;

What do you think?

[cyberman54]

It's intentional to count only local MACs, and to not count real MACs.

Reason is, that many personal devices (i.e. smartphones) use local MACs, while machine type devices (i.e..cars) use real MACs.

Tests did show, that estimation of how many humans are around, work better by counting local MACs only.

[commeco]

Oh, interesting. I didn't know that. But if there are devices for let's say several hours and they always change their mac address before scanning access points, the counting values must be wrong. Or?

The topic is old, but maybe that is also the problem here:
dbinfrago/libpax#17
If the BLE devices did not set the 'locally' bit, they were not counted.

I give it a try with counting the real mac addresses and test it on the next event. I'll come back here. :)

Should we close the issue?