EP0004340A2 - Pseudorandom generation of orthogonal matrixes for scrambling purposes - Google Patents

Abstract

1. Pseudo random generation of orthogonal number matrices for scrambling purposes, which is characterized in that another orthoganol number matrix is generated from an orthogonal number matrix already used for the scrambling, by a pseudo random permutation of its rows or columns.

Description

The invention relates to a method for the pseudo-random generation of orthogonal number matrices. Orthogonal number matrices are among others used in procedures for encrypting speech. Such an encryption method is e.g. described in German patent specification 2523828. In accordance with this method, groups of e.g. eight samples of the unencrypted speech signal are transformed using an orthogonal 8 x 8 number matrix according to known rules of matrix transformation (see e.g. Zurmühl, Matrizen, Springer Verlag 1958) and the transformed samples are subjected to further transformation processes. In order to recover the original samples on the receiver side, a matrix transformation with the inverse matrix is carried out in a corresponding intermediate step.

In order to make unauthorized decoding more difficult, provision is made, for example, in the method according to the above-mentioned layout specification, in a read-only memory of sufficient size save several matrices and, from time to time, have a pseudo random generator select another matrix that is used to transform or reverse transform the samples. Unauthorized decryption becomes more difficult the more different matrices are available and used. However, the size of a stock consisting of different matrices is limited by the capacity of the read-only memory and the necessary address memory.

The invention is based on the object of specifying a method by means of which the largest possible number of different orthogonal matrices can be generated in a pseudo-random manner with the least possible use of memory space and made available for the transformation or back-transformation of the samples.

This object is achieved in that another matrix is generated from an orthogonal matrix already used for the encryption by pseudo-random permutation of its rows or columns.

The method according to the invention, the means for its implementation and its further developments are to be explained using an exemplary embodiment.

1 shows a simple example of an orthogonal number matrix. The matrix consists of four numbered row vectors (number series from right to left) and four column vectors (number series from top to bottom). 1 can at the same time be understood as an image of the content of a correspondingly large memory matrix of a random access memory. Permutations of the rows of the number matrix, such as swapping two rows, then correspond to swap the contents of the corresponding rows of the memory matrix. For example, FIG. 2 shows the matrix that emerged from the number matrix according to FIG. 1 by exchanging the second with the fourth line vector. Transferred to the memory matrix means that the content of the second memory line has been interchanged with the content of the fourth memory line of the read-write memory. It is important in the present context that, even after the permutation, the number matrix is an orthogonal matrix and can therefore be used for the "transformation on the transmission side and the reverse transformation on the reception side in the same way as the output matrix.

3 serves to explain an arrangement with which the method according to the invention can be carried out. The components used are a random generator 1, an address memory 3, a read-write memory 4, a changeover switch 5, a first shift register 6, a second shift register 7 and a control unit 8. All components together form a clocked system, the Clock generator and the feed lines to the clock inputs of the individual modules are not shown. The random numbers output by the pseudo random generator 1 are stored in the address memory 2. The read-write memory 4 has the form of a matrix memory and has as many memory locations as the number matrix components to be treated. The control unit 8 switches the two memories 2 and 4 from read to write operation via the connections shown. In addition, it operates the changeover switch 5 and the address counter 3 and processes information about its counter reading. The number matrix previously used for encryption is stored in memory 4. The permutation process for generating another number matrix is thus initiated tet that the random number generator 1 outputs a random number to the address memory 2 and the address counter 3 is set by the control unit 8 to the address 0. With the appropriate position of the switch 5 and a control signal from the control unit 8 to the memory 4, the result is that the content of the memory cell with the address 0 is read into the shift register 6. Then the control unit 8 gives the memory 2 and the address counter 3 the necessary commands for this to be set to the address which is stored in the address memory 2. After setting the address counter, the memory line of the memory 4 corresponding to the stored address is transferred to the shift register 6, while the previous content of the shift register 6 is passed on to the shift register 7. The changeover switch 5 is then brought into a position by the control unit 8, which causes the shift registers 6 and 7 to be connected in a ring-shaped manner. It is thereby achieved that the contents of the shift registers 6 and 7 are exchanged by the next clocks. If this exchange process is completed, the control unit immediately sets the address counter 3 to 0. The content of the shift register 6 is now read into the memory line with the address 0, while at the same time the total content of the registers 6 and 7 is cyclically shifted. If the memory line is written with the address 0, the address counter 3 is set again with the contents of the address memory 2 that have remained unchanged. Now the memory line corresponding to this address is written with the content currently contained in the shift register 6. If this memory line is also written, the following final state of the memory has been reached: The content of the memory line with the address 0 has reached the memory line with the address initially generated by the pseudo-random generator while the content of this memory line was written in the line with the address 0.

Now the control unit sets the address counter again to 0 and brings the changeover switch 5 into a position in which information can be transferred from the matrix memory 4 into the shift register 6. If the content of the memory line with address 0 is written in shift register 6, the address counter is set to 1. The content of the line with the address 1 of the memory 4 is now read into the shift register 6, the content stored there being adopted into the line with the address 1. After completion of this process, the counter status is increased again by one unit and the content of the memory line assigned to this counter status is exchanged with the content of shift register 6. The process repeats itself until all addresses have run through and the counter has returned to address 0. If the line with the address 0 has been overwritten by the content of the line with the highest address, a cyclical shifting of all line contents of the memory 4 has been achieved by the method steps described. Seen overall, the content of the memory 4 has arisen from the original content by permutation of the line contents, and this permutation consists of swapping two lines with subsequent cyclical shifting by exactly one place.

What has been said so far about the rows of the memory 4 can also be carried out with the columns with only minor modifications. It is also possible that row and column permutations are mixed.

The creation of a permutation of the line contents of the memory matrix 4 according to the described method has the advantage that a number matrix is available with only one random number, which differs from the previous number matrix as a rule in each line. For cryptological reasons, the greatest possible difference between two successive number matrices is desirable. This difference could also be achieved by repeatedly pseudo-randomly swapping two lines at a time. However, it would then be necessary for the random number generator to supply more than one address in the period in which the new matrix has to be made available for encryption purposes, that is to say it is operated at a higher clock frequency. that must. Since the random generator is already operated at the limit of its highest clock frequency in most applications, a further increase in the clock frequency is impossible. The inventive design of the method for solving the problem presented shows an easy way out of the difficulties mentioned.

Claims (2)

1. Pseudo-random generation of orthogonal number matrices for encryption purposes, characterized in that another orthogonal number matrix is generated from an orthogonal number matrix already used for encryption by pseudo-random permutation of its rows or columns. 2. The method according to claim 1, characterized in that for a pseudo-random permutation of the rows of an orthogonal number matrix their components are stored in the memory elements of a matrix memory (4) and that with a control device (8) and an adjustable address counter (3) the content of the first memory line of the matrix memory and the content of a memory line, the address of which was determined by a pseudo-random generator, first read out of the matrix memory and temporarily stored in a shift register (6,7) and then read into the matrix memory so that the contents of the two mentioned memory lines of the matrix memory are interchanged and that the content of the first memory line is then read into one of the shift registers (6) and that the content of the next memory line is exchanged with the content of the shift register, and that this exchange process with the following memories hurry is repeated until each memory line has once exchanged its content with the shift register (6).

Images