QRNG chip mass production method

Abstract

The present invention comprises a test start step for initiating a final test step, a data collection step for collecting frames of bit sequences having a length of 1024 KB, preferably 512 KB, generated by an RNG chip, and a uniformity determination step for calculating the uniformity of the bit sequences according to the following formula: The present invention relates to an RNG chip testing method comprising a comparison step for comparing a determined uniformity with a predetermined threshold, and a judgment step for determining whether the chip has passed or failed a test based on the result of the comparison step.

Description

Technology Field

[0001] The present invention relates to a method for mass-producing QRNG chips, in particular to a method for mass-producing QRNG chips in a short period of time. Background Technology

[0002] Generally, the present invention is in the context of the generation of random numbers, and in particular in the context of methods for manufacturing RNGs. Many tasks in modern science and technology utilize random numbers, including in simulations, statistical sampling, game applications, and cryptography (both classical and quantum). A good random number generator must generate a chain of bits with high entropy at a high rate. High entropy implies that no one can predict the value of a bit before it is revealed, and entropy can also be understood as randomness. This is an essential requirement in most modern cryptographic algorithms and protocols. Indeed, all commonly used cryptographic protocols, such as DSA, RSA, and the Diffie-Hellman algorithm, follow Kerckhoffs' principle, which dates back to the 19th century, and the security of a cipher must lie entirely in the key—that is, in a random sequence used as a completely unpredictable seed. Therefore, it is particularly important that the key used in the encryption algorithm is secure, and this actually requires that the key be chosen completely randomly, that is, generated randomly.

[0003] To enable the generation of such random sequences, the QRNG includes controlling it to perform this random number generation.

[0004] These chips are mass-produced, and testing is, of course, necessary to evaluate the functionality and performance of each produced chip. In particular, evaluation is carried out through various tests, and among them, the Final Test (FT), which is extremely important as it is the last test performed on the chips, determines whether the chip can be released to the market.

[0005] In fact, based on FT results, only produced chips classified as good chips can be shipped to the market.

[0006] Generally, the total test time should be as short as possible to reduce production facility occupancy time, which is closely related to test costs, and to increase production capacity. Typically, the total test time should take about tens of seconds per chip.

[0007] Generally, the entropy generated by all types of random number generators follows specifications such as those defined in standards by the National Institute of Standards and Technology (NIST) or the British Standards Institution (BSI).

[0008] Figure 1 illustrates the entropy source model of NIST SP800-90B.

[0009] NIST SP800-90B includes two types of tests that examine a collection of random variables: independent and equally distributed IID tests and non-IID tests. These tests also evaluate the quality of the entropy of the chip embedded in the RNG, which is generally referred to as "randomness."

[0010] Some users request that non-IID test results be performed with very long RNG sample lengths, that is, since the test sequence represents a performance measure of the generated entropy source, it must be larger than 10 MB.

[0011] Generally, since all test cases defined in non-IID testing consist of 17 distinct test cases, executing these test cases takes a long time. Furthermore, generating and testing long data samples takes time. For example, at NIST, a test case requires 1 to 2 hours, so 10,000 to 20,000 hours of testing are required for mass production of 10,000 chips.

[0012] Therefore, according to conventional technology, there is no typical method for valuable inspection / evaluation of the quality of entropy sources in mass production.

[0013] Since most RNGs aim to provide random numbers as the final output, a common method for evaluating the quality of random numbers is based on FIPS (Federal Information Processing Standards).

[0014] However, recent QRNGs have been designed to provide entropy sources rather than random numbers. Therefore, a new method is needed to test the quality of the entropy source.

[0015] For example, in NIST SP800-90B, non-IID tests consist of 17 test cases, and "randomness" tests are performed to evaluate the performance of entropy sources. In practice, many viable noise sources do not generate IID outputs. Therefore, for non-IID data, the estimators below must be calculated from the outputs of noise sources and the outputs of arbitrary conditioning components, and the minimum of all estimates is taken as the entropy evaluation of the entropy source for the recommended tests.

[0016] Non-IID Test:

[0017] H-bit = min(MCV bit, Collision., Markov bit, Compression, T-Tuple bit, LRS bit, Multi-MCW bit, Lag bit, Multi Markov bit, LZ78Y bit)

[0018] H-org = min(MCV org, T-tuple org, LRS org, Multi-MCW org, Rack org, Multi-Markov org, LZ78Y org)

[0019] The above estimators are the Most Common Value Estimate, Collision Estimate, Markov Estimate, Compression Estimate, t-Tuple Estimate, LRS (Longest Repeated Substring) Estimate, Multi Most Common in Window Prediction Estimate, Lag Prediction Estimate, MultiMMC Prediction Estimate, and LZ78Y Prediction Estimate.

[0020] In this test, a final minimum entropy is derived in the range of 0 (low / bad) to 1 (high / best).

[0021] H-min = min (H-bit, H-org / 8)

[0022] Below are examples of test results:

[0023]

[0024]

[0025] Here, ChipId is the identity of the tested chip, and iteration is the number of test times. However, as previously mentioned, these types of test cases are not suitable for the FT process for mass production due to time constraints. Therefore, an entropy source evaluation / inspection method is required to be adapted for mass production in order to reduce test time. Accordingly, the objective of the present invention is to drastically reduce the test time of mass-produced RNGs. In this regard, the main objective of the present invention is to solve the aforementioned problems and, in particular, to provide a method for producing random number generation chips that includes an entropy source test method to reduce the total test time to be adapted for mass production.

[0026] The solution for the purpose of time reduction in the present invention is achieved by reducing the length of test data and improving the accuracy of the tests. In particular, the present invention uses a block uniformity method that tests only a portion of sample data, referred to as blocks, output by a single chip, instead of the entire data sequence. Generally, the length of the tested portion is 512 KB to 1024 KB instead of the previously mentioned 10 MB.

[0027] This block uniformity method can be repeated for each individual chip in production to evaluate whether these chips are good.

[0028] In addition to these block uniformity methods, when sample data is not long enough, the variation in block uniformity increases, and to cope with this variation, an alternative or complementary method called the pixel uniformity method is proposed.

[0029] A first aspect of the present invention comprises a test start step for initiating a final test step, a data collection step for collecting frames of bit sequences having a length of 1024 KB, preferably 512 KB, generated by an RNG chip, and a uniformity determination step for calculating the uniformity of the bit sequences according to the following formula:

[0030]

[0031] The present invention relates to an RNG chip testing method comprising a comparison step for comparing a determined uniformity with a predetermined threshold, and a judgment step for determining whether the chip has passed or failed a test based on the result of the comparison step.

[0032] Preferably, the RNG chip test method further includes a chip sample management step consisting of discarding the chip sample upon test failure.

[0033] Advantageously, the data collection step includes collecting frames of bit sequences having a length of 1024 KB, preferably 512 KB, for each of the 64 pixels of the chip.

[0034] According to a preferred embodiment of the present invention, the uniformity determination step includes the step of independently calculating the uniformity of all 64 pixels and calculating the average uniformity of all 64 pixels.

[0035] Advantageously, the comparison step compares the average uniformity of all 64 pixels with a predetermined threshold and discards chips whose average uniformity is lower than the threshold.

[0036] According to a preferred embodiment of the present invention, the RNG chip test method further includes a peak uniformity measurement step of calculating minimum / maximum peak values ​​for each pixel.

[0037] Preferably, the RNG chip test method further includes a compensation step executed only for chips that have passed the comparison step, wherein the peak uniformity of each pixel is checked.

[0038] According to a preferred embodiment of the present invention, the determination step determines whether all pixels of the chip pass the compensation step and the chip is considered a good sample, or whether at least one pixel of the chip fails the compensation step and the chip is considered a bad sample and should be discarded.

[0039] A second aspect of the present invention relates to a method for manufacturing an RNG chip comprising the RNG chip testing method of the first aspect of the present invention.

[0040] Preferably, the RNG chip manufacturing method is included in the mass manufacturing process.

[0041] A third aspect of the present invention relates to a random number generator comprising a chip manufactured by the RNG chip manufacturing method of the second aspect of the present invention. Brief explanation of the drawing

[0042] Further specific advantages and features of the present invention will become more apparent from the following non-limiting description of at least one embodiment of the present invention with reference to the accompanying drawings. - Figure 1 shows the entropy source model of NIST SP800-90B. - Figure 2 schematically illustrates the sequence of the final test process using the block uniformity method of the present invention. - Figure 3 schematically illustrates data collection for the block uniformity method of the present invention. - Figure 4 schematically shows examples of block uniformity method results. - Figure 5 schematically illustrates the sequence of the final test process using the pixel uniformity method of the present invention. - Figure 6 schematically illustrates the collection of data from each pixel for the pixel uniformity method of the present invention. Specific details for implementing the invention

[0043] This detailed description is intended to illustrate the invention in a non-limiting manner, as any feature of an embodiment may be combined advantageously with any other feature of a different embodiment.

[0044] FIG. 2 illustrates a first aspect of the present invention, which is a method for calculating uniformity in a QRNG chip. According to this method, referred to as the block uniformity method, data from a tested RNG or QRNG is collected in a predetermined length from a CMOS image sensor pixel, as shown in FIG. 3. Here, the term RNG defines any type of random number generator, including a quantum random number generator. Additionally, the term RNG may independently refer to the RNG or RNG chip as a whole, depending on the context.

[0045] As explained above, in actual scenarios, during mass production, it is impossible to run all non-IID tests to classify whether a sample is good and whether it can be released to the market due to time constraints.

[0046] For this reason, the present invention relates to an RNG chip test method comprising the step of starting a final test, followed by the step of collecting a frame of bit sequences having a length of data, preferably 1024 KB, preferably 512 KB. The RNG chip test method includes a block uniformity step as illustrated in FIG. 2, which solves the technical problem of calculating the uniformity of a bit sequence output by an RNG according to the following formula.

[0047]

[0048] Therefore, uniformity indicates how evenly data is distributed in bit units, and this formula is applied considering that one pixel has 2 bits and presents four outputs: "00", "01", "10", and "11". For example, among 400 pixel data, "00" appears 101 times; "01" appears 103 times; "10" appears 97 times; and "11" appears 99 times; then, the above formula is applied as follows:

[0049] Uniformity = (103-97) / (2x100) X 100 = 3 %

[0050] The calculated uniformity indicates the degree to which the collected data is uniformly spread across a dynamic range. This characteristic is a fundamental requirement for randomness. An additional advantage is that uniformity is more easily applied to classifying good samples during mass production. In practice, to verify chips that provide an entropy source, tests provided by NIST, such as the non-iid and iid tests described above, are performed, but they are time-consuming and lead to reduced productivity. Therefore, there are several advantages for mass production.

[0051] Once the uniformity is calculated, it is compared with a cutoff criterion / value. The cutoff criterion should be set to the upper limit of the dynamic range of block uniformity. However, as shown in Fig. 4, block uniformity has large fluctuations when the data length is not sufficiently long.

[0052] For this reason, while the block uniformity method can show the overall performance of the chip, it cannot accurately show the performance of each pixel included in the chip because a single defective pixel can be masked by the overall performance.

[0053] For this reason, an alternative / complementary test method, which is the pixel uniformity method shown in Fig. 5, was also developed.

[0054] The principle of the first steps of the pixel uniformity method is somewhat similar to the block uniformity method in that it includes similar first steps consisting of a step of starting a final test, followed by a step of collecting frames of bit sequences having a length of data, preferably 1024 KB, preferably 512 KB. Here, the main difference lies in the collection step in which frames are collected for each pixel.

[0055] Generally, the RNG chip has 64 pixels, and this method consists of the step of collecting a bit sequence frame of 1024 KB, preferably 512 KB, for each pixel according to FIG. 6.

[0056] After collecting the data, a pixel uniformity test is performed for each of the 64 pixels according to the same formula as above for the block uniformity test. It is considered that each of the 64 pixels has 2 bits and that each bit can be 0, 1, 2, or 3.

[0057] Basically, since the basic unit of uniformity is calculated as 2 bits, the block uniformity method and the pixel uniformity method have the same formula. However, as can be seen in Figures 3 and 6 above, the difference between these methods lies rather in whether the collected data and the data are calculated using the entire data or using data at the individual pixel level.

[0058] At this point, the two processes start independently:

[0059] - The first process is a linear cutoff step in which the average uniformity of all 64 pixels is calculated and compared with a threshold / cutoff criterion / value to discard the chip if the average uniformity fails the linear cutoff test, and

[0060] - The second process is measuring the minimum / maximum peak values ​​for each pixel.

[0061] At that point, if the result of the linear cutoff step is positive, that is, if the average uniformity is acceptable, a compensation step is initiated in which the peak uniformity of each pixel, that is, the maximum value among the uniformities of 64 pixels, is identified. For a 512 KB frame, the average cutoff is preferably 2.3% and the peak uniformity is preferably 10%.

[0062] This compensation step can prevent errors from occurring in a few pixels or significantly impacting overall performance due to extreme uniformity. For example, while the uniformity of 63 pixels ideally converges to 0, if the uniformity of a single pixel is greater than 10, it is not an evenly distributed source of entropy. To prevent this, the concept of compensation is a cutoff based on peak uniformity.

[0063] After the compensation step, if all pixels of the chip pass the peak uniformity, the chip is considered a good sample and accepted, or if at least one pixel of the chip fails the peak uniformity, the chip is considered a bad sample and must be discarded.

[0064] The advantage of this method is that since each pixel has unique characteristics in terms of electronic features, the data extracted from each pixel has less uniformity variation than the block uniformity method. Therefore, when only small data lengths are available in the FT process, pixel uniformity is much more effective than block uniformity in terms of yield, cost, and accuracy.

[0065] Although embodiments have been described with multiple embodiments, it is clear that many alternatives, modifications, and variations are apparent to those skilled in the art. Accordingly, the present disclosure is intended to include all such alternatives, modifications, equivalents, and variations within the scope of the present disclosure. This applies, for example, particularly to systems embedding chips and all types of hardware performing this method.

Claims

Claim 1 As an RNG chip test method, the method comprises: a test start step for initiating a final test step; a data collection step for collecting frames of bit sequences having a length of 1024 KB generated by an RNG chip; and a uniformity determination step comprising calculating the uniformity of the bit sequences according to the following formula: An RNG chip testing method comprising a comparison step in which a determined uniformity is compared with a predetermined threshold, and a judgment step in which the chip is determined whether it has passed or failed a test based on the result of the comparison step. Claim 2 An RNG chip test method according to claim 1, further comprising a chip sample management step configured to discard the chip sample if the above test fails. Claim 3 An RNG chip test method according to claim 1, wherein the data collection step comprises the step of collecting frames of bit sequences having a length of 1024 KB for each of the 64 pixels of the chip. Claim 4 In paragraph 3, the uniformity determination step comprises the step of independently calculating the uniformity of all 64 pixels and the step of calculating the average uniformity of all 64 pixels, an RNG chip test method. Claim 5 In paragraph 3, the comparison step compares the average uniformity of all 64 pixels with a predetermined threshold and discards chips whose average uniformity is lower than the threshold, an RNG chip testing method. Claim 6 An RNG chip test method comprising, in any one of claims 1 to 5, a peak uniformity measurement step for calculating minimum / maximum peak values ​​for each pixel. Claim 7 An RNG chip test method according to any one of claims 1 to 5, further comprising a compensation step to be executed only for chips that have passed the comparison step, wherein the peak uniformity of each pixel is checked. Claim 8 In claim 7, the judgment step determines whether all pixels of the chip pass the compensation step and the chip is considered a good sample, or whether at least one pixel of the chip fails the compensation step and the chip is considered a bad sample and should be discarded, in an RNG chip test method. Claim 9 A method for manufacturing an RNG chip comprising an RNG chip testing method according to any one of claims 1 to 5. Claim 10 In claim 9, the RNG chip manufacturing method is an RNG chip manufacturing method included in a mass manufacturing process. Claim 11 A random number generator comprising a chip manufactured by the RNG chip manufacturing method according to paragraph 10.

Images