[Cryptography] Best internet crypto clock
agr at me.com
Wed Oct 22 17:36:16 EDT 2014
> On Oct 21, 2014, at 1:02 PM, Bear <bear at sonic.net> wrote:
> IIRC a lot has been done to verify video and audio as having come
> from a certain moment in time or general location based on recovering
> the precise 'drift' of the omnipresent 60-cycle (or 50-cycle if
> you're Australian) hum of the surrounding electrical system. While
> it's fairly precise, it's not exact, and over very widespread areas
> the exact frequency and interference patterns recovered from a video
> or audio record have been used to determine exactly when (and to some
> extent where) the record was made.
> Relevant law enforcement and Intel agencies are, yes, known to monitor
> and record the variances specifically for purposes of dating recordings
> that later may become evidence.
Presumably this would mainly apply to video or audio files. I see no reason a still image would contain a significant hum signal, as long as the exposure time is << 1/60 sec.
I wonder how hard it would be to forge one of these power line hum time signals? Per previous posts, the only case of interest is “no earlier than”, i.e. pretending a file was recorded at a notional date and time which is later than when it actually was recorded. (Time stamp authorities solve the “no later than” case.)
The simplest situation would be if the file is created in conditions where hum is minimal, say in a well shielded room or way out in the country, far from power lines. The power line hum could then be recorded at the notional time and place and added into the file.
A more interesting case would be a file that did have a hum signal. It seems to me that with some clever signal processing, the hum signal could be modified to match the desired hum. One would first adjust the notional time of the image a few milliseconds so that the recorded hum and the notional hum were close to being in phase for as long as possible (this might place a limit on how long the video file can be). One would then compute a difference signal to be added to the recording that would make it match the hum at the notional time. It might also be possible to observe the hum on the notional day for long enough to let the forger select a time interval during the day when the frequency is close to what was recorded. Doing this would minimize the amount of alteration needed and/or maximize the length of the recording.
Given limited applicability to still images, venues where archived hum might be absent (wilderness, air planes, cruse ships, war zones, etc.) and the possibility of forgery, I don’t think the hum method obviates the need for a camera with a secure clock.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cryptography