Speech security is an important aim for users of many speech communication systems. Speech scramblers are gaining widespread acceptance as a means of enhancing protection in both military and civilian applications. Therefore, there is an increasing interest in speech scramblers due to the desire for secure speech communication over existing telephone channels with standard telephone bandwidth at acceptable speech quality and reasonable cost. In this thesis, different types of scrambling systems in different domains are presented. Selected methods of scrambling in time domain, frequency domain, and two dimensional (time and frequency) are implemented using MATLAB. Subjective and objective measures using segmental spectral signal to noise ratio were used to test all implemented scrambling systems. In these measures, the residual intelligibility of scrambled speech and quality of descrambled speech were calculated and assessed. To increase the level of security, the level of residual intelligibility is deceased, and to overcome the problem of bandwidth expansion, a packet scrambling technique in time, frequency, and two dimensional domains is presented and implemented. By the same test, objective and subjective measures, each scrambling system is tested for residual intelligibility and recovered speech quality. The results show that each scrambled signal by this technique does not suffer much from bandwidth expansion, the latter verified by sending the scrambled speech files over telephone network lines from one computer to another by using two modems. The descrambler can successfully reconstruct the original speech files from the received scrambled files with acceptable (greater than40dB) speech quality, whereas the other techniques fail to reconstruct (below 10dB) the original speech