GB2094593A - Signal encryption systems - Google Patents

Abstract

There are many situations - one example is that of the police radio communications network - where it is desirable to disguise the information content of a transceived signal such as exists in speech in the 300 to 3000 Hz band. The invention provides a signal encryption and decryption system which comprises: 1) for encryption first comb filtering means, by which the signal to be encrypted may be comb filtered in such a way that the pattern of the nulls formed by the filtering process is changed in an agile, predetermined manner, the output of this first comb filtering means being the encrypted signal; and 2) for decryption second comb filtering means, by which the encrypted signal to be decrypted may be comb filtered in a fashion complementary to that effected by the first comb filtering means, the output of this second comb filtering means being the decrypted signal.

Description

SPECIFICATION Signal encryption systems This invention relates to signal encryption systems, and concerns in particular the encryption of audio, specifically voice, signals for transception over an unsecure communications link.

There are many situations where it is desirable to disguise the information content of a signal transceived - by radio or cable, say- between the transmitting apparatus and the receiving apparatus of a transceiving system. One example is that of the police radio communications network; normally, vocal messages, such as those between the police car and its control, are transceived over narrow band radio channels that can relatively easily be "overheard" by anyone in range with a suitable receiver tuned to within a few kilohertz of the correctfre- quency, and it is clearly highly desirable to employ some form of speech scrambler, or encryption device, to improve the degree of privacy. Such a device should be simple, inexpensive and effective, and should not result in any increase in the bandwidth of the radio channel.

There area number of techniques that can in principle be used to provide speech scrambling, but those suggested so far have required an increase in bandwidth, have provided results that have been relatively easy to de-scramble, or have necessitated the use of complex and expensive apparatus. The present invention seeks to provide a simple, cheap and effective scrambling - encryption - system, suitable for use in the transceiving of information which itself exists over a finite but fairly narrow frequency band (as is the case with the human voice, most of the information in which exists in frequencies in the band 300 to 3,000 Hz), which system overcomes these problems.

In one aspect, therefore, this invention provides a signal encryption and decryption system which comprises: 1) for encryption first comb filtering means, by which the signal to be encrypted may be comb filtered in such a way that the pattern ofthe nulls formed bythefiltering process is changed in an agile, predetermined manner, the output of this first comb filtering means being the encrypted signal; and 2) for decryption second comb filtering means, by which the encrypted signal to be decrypted may be comb filtered in a fishion complementary to that effected by the first comb filtering means, the output of this second comb filtering means being the decrypted signal.

The system of the invention employs two features wherein the invention may be said to reside. Firstly, it uses comb filters the filtering effect of which may be varied in an agile, predetermined fashion, and secondly it uses two such filters the effect of one of which is complementary to that of the other, It may be explained as follows:- A comb filter is a filter which filters out (or passes, depending on one's viewpoint) the signal fed to the filter at various frequencies spaced, usually regularly, across the frequency spectrum of the signal.

The effect of a typical such filter will be clear from Figure la ofthe accompanying drawings, which shows superimposed on one another the gain/frequency plot of a constant amplitude "test" signal (dotted line) input to the filter together with that of the filtered signal (solid line) output bythefilter. It will be seen that the filter has "combed out" the input signal, forming a regular sequence of nulls (frequencies at which the gain of the output signal is much reduced, so that its amplitude approaches zero) across the signal spectrum.In the invention there are used comb filters of a type such that its filtering effect is dynamicaily variable during operation- and specifically may be altered so as to change the position and spacing ofthe nulls in an agile, predetermined manner, Such filters are discussed in more detail hereinafter, together with an explanation of their effect.

The invention requires the use (for the decryption of the previously-encrypted signal) of a second comb filter complementary to the first. By "complementary" is meant that the second filter provides an output signal with a pattern of nulls and maxima (and intermediate points) which is complementary "opposite" -to that produced by the first filter. The effect of such a filter (complementary to that whose effect is shown in Figure la) is shown in and will be clearfrom Figure 1b oftheaccompanying drawings; from a comparison of the two Figures it will be seen that the two output signals are complementary (one may be regarded as an "upside down" version of the other).Where such a second filter, complementary to the first, is used after the first - thus, where the signal output from the first is then fed to the second -the output from the second will be substantially identical in form, though possibly of reduced gain (amplitude), to that of the first, for wherever the first filter reduced the signal gain by a specific factor the second "increased" it by the same factor (in practice the second filter reduces the gain by the inverse factor, and a subsequent amplification stage then brings the signal amplitude uniformly up to the original level).

The comb filters used in the invention are dynamically variable during operation - that is, their filtering effects are alterable while the filters are operating.

The invention requires that the alterations be in an agile, predetermined, manner, and the effect of such alterations is, very simply, that a signal (the speech signal of a voice communication system, say) encrypted in this way becomes very difficult to understand. The situation may perhaps be appreciated best by considering an analogy (albeit an imperfect one), namely the understanding of a visual image generated by reflection off the surface of a liquid (a pool of water, say) which is subjected to ripples (as from successive puffs of wind). There are two main factors involved in the result; firstly there is the frequency of the ripples, and secondly there is the rate of change of that frequency. Consider the pool of water. If it is flat and calm the reflection is clear and undistorted, whereas if it is gently rippled the reflection is distorted.If the ripples are regular (of constant frequency) then the eye/brain combination can still derive some understanding of the reflection, though the extent of this may depend upon the ripple frequency (a very low ripple rate is close to a flat looking at a "foggy" object- its general nature is apparent, though detail is lost; an intermediate rate causes the most difficulty). However, if the ripples are irregular (of variable frequency) then the eye/brain combination has a very considerable problem in understanding the reflection, and will generally fail (provided, again, that the mean ripple frequency is neithertoo low nortoo high, and provided that the rate at which the ripple frequency changes - is itself neither too low nor too high).

As it is for the eye/brain combination in this analogy, so it is for the ear/brain combination with a speech signal encrypted according to the invention; the encrypting comb filter generates the varying pattern of nulls (the ripples) across the signal frequency spectrum, and the rate at which the pattern is altered is chosen so as to render the encrypted signal the least intelligible.

The pattern of nulls -their positions and spacing is related to the number of nulls and the width of the signal's frequency spectrum. When encrypting a voice signal the majority of the intelligible information in which is spread across a frequency spectrum about 3.3kHz wide it has proven satisfactory to let the null spacing be from 500 to 100 Hz, the corresponding number of nulls being from about 6 to about 30. Optimum figures for speech encryption are presently thought to be 300Hz spacing and 10/11 in number.

In the invention the pattern of the comb filtergenerated nulls is varied during operation. Naturally, the agility of these changes - the rate at which the changes occur- is important. If it is very low then the ear-brain accommodates to each change long before the next occurs, while if it is very high then the overall effect is of no rate change at all. For human speech a useful basis for assessing change rates is the length of the average syllable (about 1/3 sec); if the change rate is such that there is a change at least once and not more than ten times every syllable then ear/brain confusion is satisfactorily high. The optimum change rate is presently considered to be three or four times per syllable (that is, roughly 10 to 15 changes a second).

In addition, however, the degree of change is of importance. If the change is very small then it amounts to no change at all (though, over a longer period, the overall change may be quite large, it will have occurred so slowly as to cause the ear/brain no problem), while a large change may, fortuitously, merely result in a coarser pattern which "fits" the original finer pattern. Thus, the degree of change of the null pattern should be significant, but not such as to result in null displacement by an amount corresponding to the original null separation. Moreover, the sequence of these changes should not be such as to result in the null pattern regularly returning within a short time (two or three changes) to one which "fits" with an earlier pattern.The expression "agile" is used herein to denote a pattern change fashion which is both significant (but not too much so) and not regularly periodic; an example of such a pattern could be one wherein the null spacing was increased to near, but significantly more than twice the first spacing, was then reduced to an amount just less than half the first spacing, was then increased to an amount of one and a half times the first spacing, was then reduced to an amount less than a quarter of the spacing, was then increased to an amount of one andthree-quarters the original spacing, and finally was increased to the original spacing - the cycle then being repeated.

In accordance with the invention the comb-filtergenerated null pattern is changed in an agile, predetermined manner. It is very much preferred that the changes be apparently random but nevertheless predetermined. The reason for choosing a "random" manner of change is not so much to confuse the casual eavesdropper, for almost any pattern of changes will acomplish that, but to make it difficult, if not impossible, for the "professional" with the appropriate tools at his command to decrypt the encrypted signal.However, having decided upon "random" change it follows that the change must only seem to be random - it will in fact be pseudorandom -for in the decryption stage of the invention the second comb filter not only has to be of a type complementary to the first, it has to be actually complementary, and so must generate a matching pattern of maxima (a complementary pattern of nulls) which changes in exactly the same "random" way as that of the encrypting comb filter. The causing of some event in apparently random but actually predetermined manner is commonly and conveniently done under the control of a pseudo-random number generator. Such generators are well known, and the use of one in the invention is described hereinafter.Here, however, it may briefly be explained that such a device may be set to generate a sequence of numbers the order of which appears to be random, and that once so set it will always generate the same sequence whenever it is started (though it can be forced to start at some mid point in the sequence, if preferred). Accordingly, if the encrypting comb filter is under the control of one pseudo-random number generator, while the decrypting filter is underthe control of a second such generator which is identical to the first, the two "random" sequences can easily be caused to be the same! The inventive system involves the comb filter encryption of a signal (before transmission) followed by its complementary comb filter decryption (after transmission). The pattern of the nulls formed by the encrypting comb filter is changed in a predetermined manner, so that the complementary pattern formed by the decrypting comb filter must be changed in the same manner, and clearly the two changing patterns must be synchronised with each other. A method of synchronising the two (whether the changes are pseudo-random or not), without providing any auxiliary signals (such as timing signals) which would increase the signal bandwidth, involves pre-programming the decryption stage of the receiver with a "list" of the patterns to match.

The thus pre-programmed decryption stage is then caused to compare each pattern it receives with those in the list, and to "lock on" at that point where it finds a match- i.e., when a pattern being compared corresponds to a pattern in the list, the decryption stage then initiates further internal generation of the "decrypting" complementary comb filter null pattern starting from that point in the list of such patters. In a preferred embodiment the comparison is extended to a second (and possibly to a third) null pattern spaced in time from the first by a preselected amount; if the second pattern matches as predicted from the list upon the basis of the original matching of the first pattern, then one can be reasonably certain that the two patterns have been correctly identified and placed in the sequence of patterns, and then the decryption stage can "lock on" as described above.

In practice, however, it is unnecessary for the receiver to hold a list of all the null patterns that can be transceived; instead it need only store one (or preferably three or four fairly widely separated in the sequence), comparing it with each received pattern in turn until a match is observed. Moreover, where the null pattern changes non-repetitively under pseudo-random control it is in practice necessary merely to look for one null point at one pre-chosen frequency (which will only occur once during each pseudo-random cycle) - though again four frequencies well spaced in time will allow both faster synchronisation and a limited degree of checking.

It will be appreciated that this procedure for attaining synchronisation requires merely a foreknowledge of the changes to be made to the null pattern.

Thus it is as suitable for use with a pseudo-random series of such changes as it is with a regular series, for in each case the series is in fact predetermined (and so can be programmed into the decryption stage) even if it appears not to be.

The system of the invention requires the use of two variable effect comb filters, each complementary to the other. Although there are various ways of constructing comb filters, that presently preferred for the invention employs an analogue delay unit of the charge-coupled, or "bucket-brigade", device type to provide a delayed version of the filter input signal which is then combined with an undelayed version (the combination can be in one of two ways, as is further explained below, depending on the result required), the clocking rate of the delay unit being varied (it is conveniently the output from a frequency divider fed by a fixed frequency oscillator and the division rate of which is controlled by the pseudo-random number generator) so as in turn to vary the filter effect in the desired "random" manner.In order for such a comb filter to have the effect illustrated by Figure 1a the delayed version is summed with the undelayed version, with suitable gain adjustment, to give the required result, while for such a filter to have the complementary effect illustrated by Figure ib the delayed version is itself the required result, but is also recursively fed back to and subtracted from the undelayed version at the input (a brief explanation of thie is given hereinafter).

A comb filter constructed using as the delay unit a bucket brigade device of the type just described generates a pattern of nulls (as in Figure la and b) in which the spacing between the nulls is inversely proportional to the delay. Using a delay unit with a delay dthe nulls (in Figure 1a, say) are positioned at frequencies proportional to 1/0.5d, 1/1.5d, 1/2.5d, 1/3.5d and soon. Thus, if the delayed is set to 0.Olsec then the null frequencies will be at 100Hz intervals across the signal spectrum (at 50,150,250,350Hz....

.) whereas if the delayd is 0.001secthen the null frequencies will be 1000Hz apart (at 500, 1500, 2500,3500Hz ).

The encryption/decryption system of the invention may be employed to encrypt and decrypt any source signal that can be fed to and through the comb filters required. It is, however, the main purpose of the invention to provide a security system for use in military or civil radio communications networks and specifically for use in the police radio network.

As such, the inventive system is interposed as an additional link at a suitable place in the radio communications equipment, the encryption portion residing in the transmitter, with the decryption portion in the receiver. In a typical case, then, the encrypter is placed between the microphone/preamp. stage and modulation/output stages of the transmitter, while the decrypter is placed between the receive/demodulate stages and postamp./loudspeaker stage of the receiver.

The invention extends, of course, to a transceiving system whenever employing a signal encryption and decryption system as described and claimed herein.

The invention is now described, though only by way of illustration, with reference to the accompanying drawings in which: Figures la andb show gain/frequency curves for two complementary comb filters useful in the encryption system of the invention; Figure2 shows a schematic circuit of a radio transceiving system employing the encryption system of the invention; Figures 3a andb show schematic circuits for two complementary comb filters having the effect illustrated in Figures la and 1b; Figure 4 shows a sequence of signal plots illustrating the operation of the comb circuit of Figures 3b; Figure 5 shows a schematic circuit for the variable delay unit illustrated in Figures 3a and 3b; and Figures 6a andb show in schematic form more practical circuits analogous to those of Figures 3a and 3b.

Figures la andb have already been described hereinbefore, and need no further discussion.

The schematic circuits of Figures 2, 3a andb and 5 need no description per Se, though the following simplified explanation with reference to Figure 4 of the operation of the circuits of Figures 3a andb may prove useful.

Dealing, first, with the circuit of Figure 3a it is easy to see, if the source signal is a sine wave (varying sinusoidally in amplitude with time as shown by 40 in Figure 4), following an initial D.C. signal of zero amplitude, that if the delay effected by the delay unit is, say, 60', then the output of the delay unit will be an identical (but delayed) simple sine wave (41), and when this delayed version is "added" to the source version to give the filter's output the latter is merely a combination of the two (pictured, 42, for conveni ence as though they simply overlapped, though in practice they are actually added together). Note that for the first 60"the combination signal 42 consists solely of the source signal 40.

Turning, now, to the Figure 3b circuit, when the combination signal 42 is input to this second comb filter it firstly has the output of that circuit subtracted from it - but, of course, initially there is no such output (or, rather, the output has zero value), so the combination 42 enters the delay unit unchanged, and is output therefrom as a delayed version of itself (43) in exactly the same way, and by exactly the same amount, as source signal 40 was delayed passing through the Figure 3a circuit delay unit. Note, however, that the first 600 of this delayed combination signal 43 is derived by simply delaying by 60 the source signal 40.Thus, as the second 60 of the combination signal 42 is input to the Figure Sb circuit delay unit, so the output 43 of that unit, which output consists solely of the first 60 delayed version of the source signal 40, is first subtracted therefrom.

Accordingly, the signal actually input to the delay unit is simply the original source signal 40, which passes through the delay unit to emerge delayed by 60". Naturally, this delaying/subtraction process continues the whole time while the combination signal 42 is input to the Figure Sb comb filter, and naturally the whole time the filter outputs (43) what is simply a delayed version of the source signal 40.

Figures 6a andb illustrate more practical forms of circuit that can be employed as the required comb filters of Figure 2.

The encrypting unit is shown in Figure 6a, the decrypting unit in Figure Sb. In each case the unit may conveniently be split into a digital part and an analogue part.

The encryption unit The analogue part comprises a unity gain buffer amplifier (60) with a 600t input impedance (to give a degree of isolation) and a low (about 100t) output resistance, an analogue delay line (61) and a unity gain adder (62). The analogue delay line 61 is a bucket brigade device having N buckets, and the delay provided is given by the time a unit input impulse takes to traverse all N buckets driven by the clock inputck.The delaydis given by the expression N/fck so that for a clock frequency (fck) at 100kHz and with N =512 the delay is 5120 microseconds.

A bucket brigade delay line is a sampled data device, and the analogue output contains consider able clock frequency components and their harmon ics. A variable gain Low Pass filter (63) compensates for the gain or loss of the delay line at audio frequen cies, and also removes the clock frequency compo nents.

The whole analogue part forms a comb filter with a response which has nulls at frequencies of 1/0.5d, 1/1.5d, 1/2.5d and soon, and maxima at frequencies of 1/d, 1/2d, 1/3d and so on. The limits over which the delay unit functions are 1 to toms, giving a first null in the range 50 to 500Hz.

The frequencies of main interest for voice transmission lie in the range 300 to 3300 Hz. The comb filter operates in this range, but its characteristics are randomly altered by altering the bucket brigade delay line's clock frequency. This function is achieved by the digital part of the unit.

In the digital part the crystal oscillator (64) contains a stable crystal oscillator, a counter, and decoders. The block has three outputs: a timing signal TMG (a high frequency square wave at Fxtai/k), a clock signal ck (a lower frequency square wave at FXta,/N - -fCk=20Hz) and a gate signal GTE. GTE is a pulse which becomes "true" at crystal-controlled regular intervals and remains "false" between these intervals (in this example GTE becomes "true" every 204.75 seconds for a period of 512 ck pulses).

The sync and code blocks (65 and 66) are sync and code word generators which together form a pseudo random number generator (PRNG) consisting of a 12-bit serial-in parallel-out shift register and three exclusive "or" gates. If the parallel outputs of the register are labelled Q1-Q1 2 from the input end the exor gates are connected to the register's input so that Q3Q12=A; Q3 1tQ9=B; then the input is AB.

This gives a maximum sequence PRNG where there are 212~1 =4095 individual unique states present in the register's parallel outputs. The PRNG is clocked by the signal ck (which becomes true every 50ms); thus the cycle of 4095 states repeats every 204.75 secs, and each PRNG state lasts for 50ms.

Every 204.75 secs the GTE signal becomes true, and the Gate block (67), which is a simple gated M-bit serial-to-parallel shift register, routes a particular PRNG code to a programmable frequency divider (68).

The programmage frequency divider 68 comprises an M-bit frequency divider clocked by the timing signal TMG. Its division ratio is set by data fed from the Gate block 68. The M-bit code output from Gate block 68 sets the maximum count so that the divider counts from zero to the maximum count. This has the effect of outputting a clock signal ck which is FTMG/count max., and this in turn causes the delay unitto give a delayof512x1/fck.

The decryption unit This unit has many parts in common with the encryption unit, and those parts that are common are numbered similarly.

Again, the unit may be split into a digital part and an analogue part.

The encrypted signal enters the analogue part via a buffer (70), and is fed to a unity gain subtractor (71) where it is filtered by a complementary comb filter made out of a subtractor 71, a delay line 61, and a low pass filter 63. The decrypted signal exits the unit via a unity gain buffer (72).

The digital unit is similar two that of Figure 6a but with the addition of a synchronising input Lk to the sync. and code word generators 65 and 66. The Lk input, the effect of which is to reset the PRNG to some predetermined part of the PRNG sequence, is obtained from a code comparator (75) which is driven from a parallel-to-serial converter (74) which in turn gets its inputs from four matched filters and their comparators (75).

The filter block 75 contains four band-pass filters each of which takes an input from the buffer 70, and is set to accept one of four specific frequencies Fa, Fb, Fc and Fd. Each filter feeds a voltage comparator and a schmitt trigger, so that if one of the preset frequencies - Fa, say - is present the schmitt trigger emits a logic "1" (if this particular frequency is not present the logic signal remains at "0"). The four logic outputs L, Lb, Lc and Ld are fed into the converter 74 and thence into the code comparator 73 (which consists of a set of counters and a number of exor gates). The result is a locking signal Lk if Fa, Fb, Fc and Fd are present in a predetermined sequence and at a predetermined code, and the PRNG is then reset to a predetermined code, and thereafter then runs in synchronism with that in the encrypting unit.

Claims (11)

1. A signal encryption and decryption system which comprises: 1 ) for ecryption first comb filtering means, by which the signal to be encrypted may be comb filtered in such a way that the pattern of the nulls formed by the filtering process is changed in an agile, predetermined manner, the output of this first comb filtering means being the encrypted signal; and 2) for decryption second comb filtering means, by which the encrypted signal to be decrypted may be comb filtered in a fashion complementary to that effected by the first comb filtering means, the output of this second comb filtering means being the decrypted signal.

2. A system as claimed in claim 1 which is for the encryption of a voice signal the majority of the intelligible information in which is spread across a frequency spectrum about 3.3kHz wide, wherein the null spacing is from 500 to 100 Hz, the corresponding number of nulls being from 6 to 30.

3. A system as claimed in either of the preceding claims, wherein the rate of change of the pattern of the comb filter-generated nulls during operation is such that there is a change at least three times and not more than thirty times every second.

4. A system as claimed in any of the preceding claims, wherein the degree of change of the null pattern is significant, but not such as to result in null displacement by an amount corresponding to the original null separation, and is not such as to result in the null pattern regularly returning within a short time (two or three changes) to one which "fits" with an earlier pattern.

5. A system as claimed in any of the preceding claims, wherein the comb-filter-generated null pattern is changed in an agile, apparently random but nevertheless predetermined manner.

6. A system as claimed in claim 5, wherein the pattern changes are effected underthe control of a pseudo-random number generator.

7. A system as claimed in any of the preceding claims, wherein, in order that in operation the null pattern formed by the complementary decrypting comb filter may be changed in the same manner as that of the encrypting comb filter, so that the two patterns are synchronised with each other, there is provided, associated with the decrypting comb filter, comparison means pre-programmable with a "list" of the patterns to match, this comparison means acting to compare each received null pattern and, when a pattern being compared corresponds to a pattern in the list, initiating further internal generation of the decrypting complementary null pattern starting from that point.

8. A system as claimed in any of the preceding claims, wherein each comb filter employs an analogue delay unit of the charge-coupled, or "bucket-brigade", device type so as in operation to provide a delayed version of the filter input signal which is then combined with an undelayed version, the clocking rate of the delay unit being varied so as in turn to vary the filter effect in the desired manner.

9. A system as claimed in claim 8, wherein in operation in the encrypting filter the delayed version is summed with the undelayed version, while in the decrypting filter the delayed version is itself the required result, but is also recursively fed back to and subtracted from the undelayed version at the input.

10. A signal encryption and decryption system as claimed in any of the preceding claims and substantially as described hereinbefore.

11. A transceiving system whenever employing a signal encryption and decryption system as claimed in any of the preceding claims.