Scalabiblity of witnessing

[dbosk]

How scalable is witnessing? If there is one witness that must communicate with each of one million protesters, will that work?

[dbosk]

We analyse the worst case: We have one trusted witness that must witness every protester.

Let's assume that the range of the DB protocol is 200 m (in ref to 3db-access). Then the witness covers 200^2 \pi = 125 600 m^2 when at a fixed location.

Assume the measurements the police use in Seoul: 3.3 m^2 can keep 6 people sitting or 10 people when standing. We assume 10 people per 3.3 m^2 since then the witness must run the protocol more. This yields 125 600 / 3.3 * 10 = 380 606 people in that area.

Assume that the DB protocol takes 1 s to run. Then the witness would have witnessed everyone in the area after 105 h = 4.4 days. The other way around, say that the witness is at the location during a day, i.e. 8 hours, then the witness can witness 28800 protesters.

Then we would need (almost) 35 witnesses to cover 1 000 000 protesters over a day (8h).

[dbosk]

3db-access says < 1 ms to run and > 120 m line of sight, then the numbers would be:

• Area: 120^2 \pi = 45 216 m^2
• People in area: 45216 / 3.3 * 10 = 137 018
• Everyone in the area: 2.5 minutes
• Per day (8 h): 28 800 000 people

Furthermore, they say on the same page that the energy consumption is 10 uJ per measurement. That makes it 0.0001 kWh over a day, which is negligible.

[dbosk]

According to our reference [29], it took a Samsung Galaxy S i9000 around 153 ms to run Idemix back in 2014. Using that number gives us:

• Everyone in the area: 5.8 h
• Per day (8h): 188 235 protesters
• Witnesses/million/8h: 5.3 witnesses/1 000 000 protesters over 8 h

[dbosk]

Summarized in rebuttal. Add to paper?