Abstract: A system and method are disclosed for encrypting a voice signal. The input signal is filtered to provide a signal in a first path with components in a higher frequency band and a signal in a second path with components in a lower frequency band. Analog-to-digital conversion is carried out at a rate such that the frequency composition of the higher frequency path is inverted and shifted into the lower frequency band. Signals from the two paths are intermixed and converted to analog form at a rate greater than twice the rate of analog-to-digital conversion, so as to provide a randomized sequence of time segments of signals from the two paths, interspersed with gaps.