An easy version of the method for using chaos to encrypt an audio message described in Strogatz's Nonlinear dynamics and chaos, 1994

encryptWithChaos.m

%% Encrypt with chaos
% An easy version of the method for using chaos to encrypt an audio message
% described in Strogatz's Nonlinear dynamics and chaos, 1994
%
% More information at: <http://mappingignorance.org/2016/05/09/the-sound-of-chaos/>

%% Load the original sound
load handel;

sampling_freq = Fs;
signal = y';
length = size(signal, 2);

%% Harvest chaos from the Lorenz equation
% The Lorenz equation is the classical example of deterministic dynamical
% system giving rise to chaos. The equation reads:
% 
% $$\dot x = d \left( y - x \right)$$
%
% $$\dot y = r x - y - e x z$$
%
% $$\dot z = -bz + e x y $$
%
% With the following set of parameters
%
% $$d = 10, r = 28, b = \frac{8}{3}, e = 1$$
%
% we will find chaos for almost all initial conditions, being the only
% exception the coordinate origin (an unstable steady state).
%
% So now we can pose the problem in Matlab code

d = 10;
r = 28;
b = 8/3;
e = 1;

fun = @(t,x) [d*(x(2) - x(1));
              r*x(1) - x(2) - e.*x(1).*x(3);
              -b*x(3) + e.*x(1).*x(2)];

%% Solve numerically
% The Lorenz equation is not analytically solvable, so we will solve it
% numerically for a given initial condition
x0 = [0; 1; 0]; % Set initial condition
tSpan = [0 300]; % Set time span
sol = ode45(fun, tSpan, x0); % Create solver object

% Resample with the desired time step
t = linspace(tSpan(1), tSpan(end), length); % We want a vector with the same...
x = deval(sol, t); % ... length as the signal

%% Extract a one-dimensional time series
% An easy way to extract a one-dimensional time series out of the Lorenz
% attractor is focusing only on one coordinate. Z in our case
timeSeriesX = x(1,:);
timeSeriesZNormalized = timeSeriesX./max(abs(timeSeriesX));

%% Use it as encoding key
strength = 5000;
key = strength*timeSeriesZNormalized;

coded_signal = signal + key;
decoded_signal = coded_signal - key;

%% Plot the signals
figure;

subplot(4,1,1); 
plot(t, signal, 'Color', 'b');
title('Signal');

subplot(4,1,2);
plot(t, key, 'Color', 'r');
title('Key');

subplot(4,1,3);
plot(t, coded_signal, 'Color', 'g');
title('Coded signal (Signal + Key)');

subplot(4,1,4);
plot(t, decoded_signal, 'Color', 'b');
title('Recovered signal (Coded - Key)');

%% Export to audio files
%
% The _sonification_ process follows this steps:
% 
% * Normalize the signal to the range [-1,1]
% * Pick a sampling frequency
% * Export as _.wav_
%

% Normalization
codedNormalized = coded_signal./max(abs(coded_signal));
decodedNormalized = decoded_signal./max(abs(decoded_signal));

% Exporting
audiowrite('signal.wav', signal, sampling_freq);
audiowrite('coded.wav', codedNormalized, sampling_freq);
audiowrite('decoded.wav', decodedNormalized, sampling_freq);

%% For direct playback use
% 
%   soundsc(decodedNormalized, sampling_freq);
% 

%% Credits
% Pablo Rodríguez-Sánchez 
% 
% <https://sites.google.com/site/pablorodriguezsanchez/>
%
% April 2016