Fig. 5

Comparison of the Euclidean distance (x axis) of the original data and the normalized Hamming distance of the embedded data (y axis). Left panel The security parameter Δ influences the position of the “knee” (m = 8192). Right panel An increase in the embedding dimension m decreases scattering and thus the correlation between the original and the embedded distance in the linear part of the curve (Δ = 15)