Method for authentication

Problem underlying the invention

0001A client should authenticate itself to a server without transmitting its password in plain text. Even a password encrypted with the same key can be intercepted and used by unauthorized persons.

Previous solution to the problem

0002With the conventional method, a randomly signed challenge text (signed challege) encrypted by the server is requested and sent by the client. The client decrypts it, forms a digest and signs and encrypts it again (signed digest). The server checks the digests for agreement. This process is very complex.

Possible other solution to the problem

0003Similar to a TAN list (one time pad), you could exchange a (finite) list of keys (keys) before the first authentication by other means (e.g. paper), which should only be used once. However, this list would have to be saved or the keys written down by the user (vulnerable or cumbersome and prone to errors).

Solution of the problem according to the invention

0004The invention is explained in more detail below with the aid of the drawing, which comprises a figure.

0005Both sides (client and server) agree at the beginning (as part of a seed) with the same starting value and the same algorithm for generating keys.

0006This enables the client and server to generate the same sequence of keys independently of each other. A sequence that can be generated in this way is also referred to as PRBS (Pseudo Random Binary Sequence). A PRBS can be generated, for example, using an LFSR (Linear Feedback Shift Register).

0007However, PRBS are strictly deterministic. Therefore, the procedure must be determined after a small number of results (keys) and an attacker then knows all the other keys. In addition, an exact repetition of the key sequence occurs after a finite number of keys (period of the LFSR).

0008However, if disturbances are added to the process at regular intervals (ie in the formation of the keys using the LFSR), the result is neither periodic nor deterministic.

0009The disturbances in the process are achieved as follows (see also the figure).

0010First, the client and server use the LFSR to form a first key from the start value (start key), which is referred to below as the epsilon key ε for better distinction. The epsilon key is not used for authentication because it is vulnerable.

0011Next, only the client generates a second key, which it sends to the server. Only this second key is used for authentication. In order to understand the formation of the second key, which differs from the formation with the help of the LFSR, a small digression into the metric is made.

Digression:

0012Metric means how the distance (delta) of two points is defined. Applied to numbers that are stored in bits by the computer, in the simplest case this means the difference between two numbers. The difference between two binary numbers depends on the value of the individual bits of the numbers. Normally the value of the bits is 0 (2nd⁰) to 31 (2nd³¹) fixed (old metric):<tables id="tabl0001" num="0001"><img file="EP2057778B1_D0001.tif" /></tables>

0013The distance between the numbers 3040593782 and 2503657302 has the value 536936480 according to the old metric. If a new metric is now defined for different bits by the assignment or appointment of new values, the numbers mentioned have a different distance. For example, bit 16 gets the new significance of bit 0 (2nd⁰), Bit 29 the new significance of bit 1 (2nd¹) and bit 5 the new significance of bit 2 (2nd²), then the two numbers shown above are only 7 apart. The order of the other bits is irrelevant, but not the state (0 or 1) of the bits. If you look at a delta of up to 7, the new metric shows the number 3040593782 in the delta of 2503657302, the number 2503657303, however, is more than 7 away from 2503657302 because one bit differs that has a higher value than 2.

0014In the invention (as part of the seed), an appointment is made between the client and the server via the new metric in that those bits (so-called interference bits) are agreed that have a different value, different from their transmission order.

0015Assume that 3040593782 is the first key (epsilon key) generated by the client and server. The client then forms the next key (second key), the distance (delta) of which, according to the new metric, lies within a predetermined value to the epsilon key. This key is referred to below as the delta key δ. If the client generates the number 478651654 as the next key, it is a delta key, since this number is in the delta of the epsilon key 3040593782.

0016Since only the client and the server know the new metric changed according to the position of the interference bits, the server can thus authenticate the client on the basis of the second key received. If the second key sent by the client is within the delta, the client is authenticated. This key will be used as the new seed on both sides and another shift in the LFSR will be the new first key (new epsilon key).

0017In the example shown above, only 32 bits are shown. In reality, the numbers are much larger, e.g. 2048 or 4096 bits with 8 or 16 interference bits. Operations other than the difference can also be used to determine the distance.

0018If n is the number of interference bits, then they all form 2ⁿ-1 delta keys the delta, i.e. the area around the epsilon with a known metric, which in turn is known to both sides due to the position of the interference bits and which the attacker cannot determine.

0019Advantages of the new solution: At the beginning, as part of the seed, a start key has to be exchanged using the conventional method. Thereafter, only one key from the new Delta has to be sent for each new registration.