[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
Re: [tor-bugs] #23523 [Core Tor/Tor]: Handle extreme values better in add_laplace_noise()
#23523: Handle extreme values better in add_laplace_noise()
-------------------------------------------------+-------------------------
Reporter: teor | Owner: teor
Type: defect | Status:
| needs_revision
Priority: Medium | Milestone: Tor:
| 0.3.3.x-final
Component: Core Tor/Tor | Version: Tor:
| 0.2.6.2-alpha
Severity: Normal | Resolution:
Keywords: tor-relay, privcount, review- | Actual Points:
group-24 |
Parent ID: #23061 | Points: 0.5
Reviewer: catalyst | Sponsor:
| SponsorQ
-------------------------------------------------+-------------------------
Comment (by teor):
Also, a close reading of the proof in section 5.2 of the paper [0] shows
that it only applies up to 2^46^, not 2^53^.
So we need to define a constant and use that instead.
And I think we should error out on platforms that don't have mantissas of
2^53^.
(How do we report numbers greater than 2^46^?
We increase the sensitivity of the signal.
Practically, this means we multiply the final noise by a scaling factor
2^N^, clear the N lowest bits of the signal, and add them together.
We have to clear the signal bits, otherwise signals of 0 are biased,
because they only occur half as often.
This is what we tried to do by rounding the hidden service counts to the
nearest figure, but it's not implemented correctly.)
[0]:
https://pdfs.semanticscholar.org/2f2b/7a0d5000a31f7f0713a3d20919f9703c9876.pdf
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/23523#comment:12>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online
_______________________________________________
tor-bugs mailing list
tor-bugs@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-bugs