What does scrambling mean in digital communication

Voice obfuscation

Sun Jun 10 18:06:34 2001


Anyone who wants to disguise telephone or radio communication can take a variety of paths. The oldest types of devices for speech obfuscation are the "scrambler". They encrypt what is spoken by messing up the order of what is spoken ("time domain scrambling"), or by changing the frequencies that make up human speech ("frequency domain scrambling"). After "scrambling" the encrypted message can no longer be understood, but it is possible to recognize a human voice through the strangely deformed noises.

It is also possible to digitize the language first (convert it to zeros and ones) and to encrypt the resulting bit rows. The encrypted message must then be converted back into an audio signal that is suitable for being sent via a telephone or a transmitter. This procedure may sound awkward and complicated, but it has several advantages. In contrast to the language itself, the bits can be processed with more complex encryption recipes.

"Scrambling" methods only make sense if they work in "real time", ie the encrypted messages are sent so quickly that direct communication is still possible.

This "scrambling" form divides the spoken text into "blocks" approximately every half a second. The device saves these "blocks", divides them into smaller pieces and mixes them up according to a certain pattern, and the shattered "blocks" are sent. This is comparable to encryption using the permutation method: it is not the characters / signals themselves that are changed, but only their order. The reverse procedure is used for decryption.

Because a piece of "speech" always has to be temporarily stored, there is always a delay of about half a second on the sending and receiving side, which means that the communication delay is about one second. That is not much, but it does mean that those who communicate with each other need to be patient. Talking confused or at the same time is not advisable.

If half a second is subdivided 15 times, the number of mixing options is mathematically quite large, in any case much too large to be able to simply try to put the original message back together again. However, as I said, it is possible to hear in the tone sequence of the beeper and snippets of speech whether e.g. a man or a woman is speaking. Persistent eavesdroppers can surely recognize the individual interlocutors over time.

Speech consists of sound waves of different frequencies. During frequency conversion, the frequency that makes up the language is processed. Each frequency is changed to another. In the somewhat older systems, this was always done according to a fixed "key" (conversion frequency), but this turned out to be easy to crack.

For more modern systems, a different key is used for each frequency of the language to be distinguished during the conversion. Lower frequencies are converted to higher and higher to lower. At the moment, systems that use constantly changing keys are mainly used. The greater the number of keys in the device, the more difficult it is to crack the system. Another great advantage is that there is no delay during communication. The principle is common for telephone and radio communication connections.

There are also devices that combine both methods. Such encryptions are correspondingly more difficult to crack. However, these devices also have the disadvantage of the former method: there is a delay of about one second. Until recently it was only possible for private individuals to buy a limited number of simple scramblers. However, these do not guarantee real security for a long time. However, more modern scramblers and the first digital voice encryption systems have recently become commercially available.

With the most modern speech encryption techniques, no deformed speech is sent after encryption, but a signal that contains bits. Zeros and ones are displayed with distinguishable beeps or tones. Until a few years ago, this whole process - digitizing, encrypting and converting the bits into a suitable signal (modem) - still led to problems. On the one hand, as a result of digitization, the number of bits (too) was obtained, and on the other hand it was not possible to send these bits in real time.11.1 There are now methods of audio digitization that generate fewer bits. Modem transmission techniques have also been optimized.11.2 The transmission speed has increased enormously. Developments that also open the way to digital speech obfuscation.

In principle, we can use the same recipes for encryption as we described earlier: DES, IDEA, a pseudo-random key or an XOR operation.11.3 Devices that use pseudorandom keys or DES are the most common. After digital voice obfuscation, only a noise can be heard and no conversation can be recognized. The American company Motorola is one of the first to bring a system onto the market that is suitable for (mobile) radio communication ("Digital Voice Protection", or DVP)11.4). Other companies such as Marconi, Ascom or Philips meanwhile also supply digital voice encryption systems with different technical levels. In addition, digital encryption systems are now also available for radio communication (by telephone, fax or modem).

Of course, with every new encryption system the question always arises as to whether a back door has been built in somewhere that enables the manufacturer (or state authorities) to overhear. Anyone who buys a finished product never really knows what it contains. If you decide to purchase such a system, British Ascom is probably still the best choice.

We haven't mentioned the price yet, hold on! Two encryption units (one on each side) quickly cost 12,000 DM. You still don't have any devices for making keys. A simple PC program with a cable to the crypto phone costs around 5000 DM.

The better voice encryption apparatus are expensive and, moreover, difficult to obtain commercially. However, you don't need to be a child prodigy to assemble an encryption system for a normal telephone with the devices that are usually already in the house. You will need a PC (for encryption), a sound card (for recording and playing back the sound), a modem (for communicating with the other side) and of course enough expertise and patience to connect the whole thing. Unfortunately, it does not (yet) provide a practical, mobile whole. Nevertheless, we would like to pass on the first tips on this area.

The first step is to convert the language into as few bits as possible. In recent years, great progress has been made in the field of digitized sound and image design (multimedia). All kinds of sound cards (audio cards) have been put on the market that convert bits into sound. These cards can be built into the PC; the required software is usually supplied with the card. A well-known example of such a sound card is the Sound Blaster. The modern cards employ fairly effective compression techniques. Such cards are available from a few hundred marks. When buying, you should make sure that the compression takes place with the help of the hardware. Cards are commercially available in which the operating information states that compression is possible, this is then sometimes done with the aid of the software, which makes the process too slow.

There are various techniques with which speech can be converted into bits, namely pulse code modulation (PCM), delta modulation (DM) or delta sigma modulation, the technique of the subband coder / vocoder (e.g. Mpeg audio Coder) and the Lineais predictive coding (e.g. LPC-Celp). The last two techniques mentioned produce the lowest number of bits per spoken second in the end, LPC would even have to manage 740 bit / sec, but the techniques are unfortunately not yet installed as standard in conventional audio cards and are therefore correspondingly expensive. Readers with electrical engineering skills can find circuit diagrams in specialist magazines and possibly solder themselves together a little.

The (cheaper) chip, which is often built into audio cards as standard and provides compression, is called DSP. The compression methods that support this chip are called AD-PCM, mu-Law and A-Law. Cards that support compression with the DSP chip are, for example, the Sound Blaster 16 MultiCD (approx. 500 DM) and the Microsoft Sound System 2.0 (approx. 460 DM).

If the speech signals have been converted into as few bits as possible, then those bits have to be encrypted again. With digital voice encryption, in principle the same encryption recipes can be used that we have already described. Now, however, the speed of the algorithm is of greater importance, which of course also partly depends on the performance of the computer used.

IDEA block encryption is most suitable for experimental purposes. This is on average twice as fast as DES and seems to be (more) secure. The software version is freely available in the original coding form, but would have to be fundamentally adapted, since the 128-bit key that IDEA uses is based on (part of) the message. This is of course not a good starting point for real-time voice encryption.

Because encryption also requires a lot of storage space, it is recommended to use at least a 386 DX with 4Mb of RAM.

A modem is required to send the encrypted voice. The most modern, but now still very expensive, modems already achieve transmission speeds of around 24,000 bps (real speed). A modem with a speed of 14000 bps and an integrated error correction mechanism is available for 300 DM. Such a modem is in principle suitable for "real time" transmissions, provided that the digitization has not resulted in too many bits. Encrypting these bits now takes up most of the time.

If you want to build a system yourself, with the current technical means you will likely get a somewhat impractical system that can only be used from a fixed location. A modem connected to the car phone cannot operate at very high speeds, and the audio cards available for laptop computers are not fast enough because they are mostly external devices.11.5

Contact: [email protected]