[back for only 1 thing] Algorithms for Cheap Shielding

Karl gmkarl at gmail.com
Thu Oct 15 04:44:01 PDT 2020


Please ask me to stop in whatever words you like, with a reference to
this request, if I talk about anything other than this.

I'd like to make a cheap device for taking ongoing measurements of the
effectiveness of a shielded enclosure.  Would you be willing to
briefly review my first prototype theory, and _concisely_ offer any
_concise_ problems or _concise_ better ideas you can think of?

> Noise source emits EM signal noise function, which at the receiver is present
> as Rn(t).  This signal oscillates in what is roughly a square wave, for example
> with a relay.
> During inactive time, receiver receives only Rb(t) the background signal.
> During active time, receiver receives Rb(t) + Rn(t) = Rr(t).
> When the signal is weak, heuristics can't differentiate between active and
> inactive time periods.  So, we separate the problem of keeping the receiver in
> sync with the emitter from that of measuring the signal, and compare the
> probability distributions of the active and inactive timeperiods to discern the
> strength of Rn(t).  Now, summation over time can be used to increase sensitivity
> if the signal is very weak.
> First, estimate the probability distribution of Rn(t) from our observations of
> the probability distributions of Rr(t) and Rb(t), possibly using e.g. the
> convolution theorem for the sum of two independent random variables, which is
> not complex.
> Then, discern mean electromagnetic properties such as relative power by solving
> from the distribution of Rn(t).

I've started pursuing this algorithm and goal a couple times over the
years, and I'd
like to try again.  Step 1 is to identify in software the precise
oscillation period of
such a waveform, similar to implementing a van eck phreaking attack.

I imagine while exploring step 1, plotting the accumulated shape of a
single cycle
visually, as if an oscilloscope had a trigger set for it.  Each time
point in the plot would
be a fully colored column, like in a spectrogram, but of the frequency
of occurrences of
different signal powers, rather than spectral decomposition.

I think such a plot would be worthwhile because it would help move
towards both this goal,
and other goals in electromagnetic security, of quickly reviewing the
details of a repeating
signal.  Such signals are common in clocked digital electronic systems.

More information about the cypherpunks mailing list