TW580797B - Random number generator using audio card - Google Patents

Abstract

A random number generator using audio card is disclosed, wherein the audio card widely installed on the personal computer or Unix workstation is used to extract the noise value in the microphone port as the primitive value of the random number sequence, and a procedure is performed to eliminate the concentrated distribution and the timing correlation in the noise signal, so as to obtain a real random number sequence with high quality. The principle of the present invention is based on the unpredictable property of the thermal noise in quantum mechanics. Thermal noise is extracted and used as the source of actual random number, and the characteristics of the hash algorithm such as scrambling/diffusion and compression in cryptography is utilized to relieve the concentrated distribution and the timing correlation in the noise value, and to greatly improve the statistical quality of the random number sequence. In addition, by using different sources of random number to form the random number sequence, the degree of difficulty in guessing the system random number can also be increased proportionally.

Description

The technical field to which the invention belongs, prior art, content, implementation, and drawings are briefly explained. [Technical field to which the invention belongs] The present invention relates to a random number generator using a sound card, and particularly to a personal computer or a Unix workstation The sound card on the card captures the bluffing sound value on the microphone port as the original value of the random number sequence, and then processes the centralized distribution and the back-and-forth correlation in the noise signal to obtain a high-quality random number generator of the random number sequence. . [Previous technology] A high-quality random number generator plays a very important role in many application systems, such as: security systems, computer games, color selection, system simulation, etc. The random numbers used in current computer systems can be divided into two categories: virtual random numbers made by software algorithms, and the data using random events on the hardware as the face of true Wei numbers. Although the chaos number can have some good performance in statistical distribution, it is essentially a predictable output. A specific initial value produces a specific virtual disorder value, and its randomness is not much better than a certain natural number sequence. Furthermore, unpredictable random numbers occupy a critical position in many application systems, such as: The keys required for cryptographic calculations should have unpredictable properties in order to ensure the security of the system. However, the random numbers generated on general computer systems are calculated by software algorithms from a given starting value. This kind of seemingly random values are arranged in order and lack randomness. At present computer efficiency It is already possible to complete the analysis and grasp the law of random numbers within a reasonable time. Random numbers that are generated randomly without any regularity to predict need to be generated by hardware. For example, the thermal noise of electronic components is used as the source of random numbers, which is the indispensability of free movement of electrons in the theory of uncertainty based on quantum mechanics. Predictive. However, how to collect the thermal noise on electronic components in a general computer environment? Continued page (When the description page of the invention is not enough, please note and use the continued page) 6 580797 It is not easy, usually using exclusive hardware The body is inserted in Rs_232 or USB, and then the secret generated by the hardware is retrieved through the driver, but this increases the cost and inconvenience of the hardware and installation. The high-quality random numbers are becoming increasingly important in computer systems. Intel Corporation's later versions of pentiumii also provide hardware-generated real Wei numbers. However, it still needs to install random number drivers and set them. There are also some inconveniences. It can be seen that there are still many shortcomings in the above-mentioned conventional technologies, and they are not a good designer. They need to be improved. In view of the various shortcomings derived from the conventional random number generation method described above, the inventor has been eager to improve and innovate. After years of painstaking research, he finally successfully developed this random number generator using a sound card. [Objective] The purpose of the present invention is to provide a random number generator using a sound effect card, which uses a sound effect card commonly installed on a personal computer or a Unix workstation to capture the noise value on the microphone as the source of the random number sequence. There is no need to install specific drivers or make any settings, which greatly improves the convenience in use and still has unpredictable quality of the hardware. In addition, in terms of removing the deviation of the threat signal, the theory and The results obtained from the experimental analysis are designed for post-processing functions to take into account the performance of the system under the best random number quality. A second object of the present invention is to provide a random number generator using a sound card, which uses unpredictable thermal noise as the original value of the random number, and then uses a carefully designed cryptographic hash algorithm to scramble the signal. The obtained real random number sequence has excellent performance on various main randomness evaluation indexes. Continued pages (note if the invention description page is insufficient, please note and use the continuation page) 7

Another object of the present invention is to provide a kind of random number generator for sound cards. A 512-bit one-way hash algorithm for post-processing operations of random number sequences not only improves the quality and quality index values of random numbers. It also eliminates the possibility of reverse cracking the original noise sequence. In addition, we read the system's high-resolution _ value Wei Si A Qianxing mask processing, the number order _ guess difficulty. The defeat number generator using a sound card that can achieve the above-mentioned object of the invention is achieved by the following principles: Unpredictability of free movement of electrons in the uncertainty principle of quantum mechanics, capturing thermal noise to obtain random random numbers Initial value; Cryptography's miscellaneous masking features such as obfuscation, diffusion, and compression to remove deviations from noise signals. In addition, the one-way nature of this hash function can also eliminate the possibility of backward calculation of the original noise value; combining several different random random number sources to form the final random number sequence can increase the random number of the system in proportion. Difficulty guessing. These principles and several experimental results and their applications are described as follows: 1. Thermal Noise (Thermal Noise or White Noise): Thermal noise is discovered by J.BJ〇hnS〇n, also known as J〇hns 〇n η〇 & This kind of noise-originated from the random movement of the electrons inside the resistive element, its strength is proportional to the absolute temperature of the resistance environment; the existence of this noise can be known from the following experiments ... Place a general resistance device without any power supply at room temperature. At this time, the resistance should be measured without any voltage value, but it can be measured with a precision electronic voltmeter instead of a DC voltmeter. To the existence of a random voltage value. This thermal noise is based on the uncertainty principle of quantum mechanics and has truly unpredictable randomness. D Continued pages (If the invention description page is insufficient, please note and use the continuation page) 580797

Thermal noise is generally expressed by the following formula: K-2 = 4kTR (f) Af 2 where “is the average squared value of the voltage drop generated by thermal noise; k is the Boltzmann constant = 1. Probability, f is the frequency of the response; r (0 is the impedance value of the optoelectronic element, generally a function of frequency; Af is the frequency bandwidth. 〇R (f) is the value of '' Let R (f) = R, Af = B, the noise of the above formula can be represented by an equivalent circuit, where the voltage source is thermal noise = 4kTRB = K ?, or electrical_ is y = 4KTB / R. 2. Whether the sound card is plugged into the microphone or not The influence of random values: Test separately: ⑴ indecent grammar; (2) plug in the microphone but no sound is input; ⑶ plug in the microphone but give a monotonous sound input; ⑷ plug in the microphone and give a rich sound input; Let ’s wait for four cases to perform random numerical analysis of people ’s losses, and learn that the signal obtained by the ridge reduction is the most random, followed by the cases of ⑶ and (2), and finally the case of ⑴. But the difference between the four is not Large, especially for the index of chi-square distribution. Therefore, the design of the present invention can ignore whether there is a microphone inserted on the sound card. This is fine, but it is necessary to detect whether the hardware of the sound card is abnormal and generate a fixed input signal, because the random number sequence obtained under this condition cannot have good random properties. 3. Hash function The effect of the input / output length on the random value: The entropy value of an random number sequence according to the net theory in information theory is: H = E [Pi * (-l〇g P〇) that accumulates every possible output The probability is multiplied by its negative log value. So you can guess whether the longer hash input value and the longer hash output value can improve the quality of the random sequence. □ Continued page (when the invention description page is insufficient, please note And use the continuation page) 9 Observe the hash function in the random number post-processing program of the present invention in: (1) different input data is produced long; (2) different output hash value length; The effect of quality. The results show that after the input data exceeds a certain length (several bytes), there is no significant effect on the quality of the random number. However, the different output hash value length will affect the entropy of the random number, the longer The higher the entropy of the hash value However, the length of the hash value has no significant effect on the Chi Square distribution value of the random number sequence. Therefore, the present invention can use a longer hash function to perform the post-processing operation of the random number sequence to obtain a better random number. Quality. Although the length of the input data can theoretically increase the randomness of random numbers, it usually does not have a significant improvement after it exceeds a certain length (different due to different operating systems and sound card settings). Effect. In practice design, the length of the input data is usually not small enough to affect the quality of the random number. Therefore, the length of the input data of the hash function should be increased as much as possible while meeting the system performance requirements. [Preferred Embodiment] The present invention uses a sound effect card to produce a high-quality random number sequence. The production process can be divided into two programs, namely, a thermal noise acquisition program and a random value post-processing program. The thermal noise acquisition program uses sound cards commonly installed on personal computers or workstations to read sounds and signals without plugging in a microphone, and determines whether the input signals are normal or not. The post-processing program depolarizes these thermal noises to obtain high-quality and unpredictable random numbers. I. Thermal noise acquisition procedure: First read the sound signal of a certain data length from the sound card, INaud = ReadMicrophone (INPUT—BUFFER—SIZE) INaud represents the noise value of the input sound card, jNpux ^ BUFFEi ^ SKE is every The length of the data read each time can be set at the time of system installation. The maximum value depends on the system's final random number of D continuation pages (if the description page of the invention is insufficient, please note and use the continuation page) 580797

The minimum value depends on the change period of the microphone signal and the length of the hash function for random post-processing. The principle for determining the length is that if the signal on the sound card does not have sufficient variation under a certain sampling length State, the length should be set to the minimum value of INPUT_BUFFER_SIZE to obtain randomness of random numbers, and this minimum value should also be greater than the length of the hash function after random numbers are processed; The speed of the number sequence is also slower, so a trade-off must be made between these two factors to determine the appropriate INPUT_BUFFER_SIZE. An application system that requires more randomness can increase the INPUTJBUFFER-SIZE.

Performers take the minimum. That is: 512 bits ^ INPUT_BUFFER SIZE $ INPUT—BUFFER—SIZE for minimum RNG speed

In order to help the user to effectively set the INPUT_BUFFER_SIZE value when installing the random number generator system, the present invention provides a microphone signal analysis tool program to facilitate finding the best INPUTJBUFFER_SIZE value during the installation process. ; This tool refers to the Entropy value of the microphone input signal, and continuously tests the Entropy value generated by different read lengths (from 64 tests to 256 bytes' in increments of 32 bytes), and then finds the INPUTJ3UFFERJSIZE, which produces the largest Entropy value under a shorter read length, is used as a parameter and written into the system's configuration profile; if the system performs too slowly under this read length setting, the user can edit this one by himself Configure the profile and reduce the value of INPUT_BUFFER_SIZE to achieve the required execution performance. When the system reads these input signals for the first time, a signal value analysis will be performed to determine whether the microphone is in a normal state and can provide random thermal noise, so as to avoid the poor quality of random numbers caused by the failure of the sound card hardware; judgment The method is based on the correlation before and after the signal. □ Continued on the next page (if the description page of the invention is insufficient, please note and use the continued page) 11 580797

The absolute value of the right correlation coefficient (Serial correlation coefficient) greater than 0.9 should be considered a hardware failure, the system returns a relative error message and terminates the execution; if the absolute value of the correlation coefficient is between 0.7 and 0.9 , Continue to read more sample signals for further discrimination. The correlation coefficient INscc = ComputeCiorTelatiaiiClaeffideiit (INaud) of the input noise value. If the absolute value of the correlation coefficient obtained by continuous sampling is between 0.7 and 0.9, the system returns an error after 5 sampling tests. Message and terminate execution. Second, post-processing of random values: The post-processing program of random numbers uses the characteristics of confusion, diffusion, and compression in hash algorithms in cryptography to eliminate the centralized distribution of noise input values and the front-back correlation, and can greatly improve Statistical quality of random numbers.

Outhash = SHA—512 (INaud)

Outhash indicates the random number output value obtained after the 512-bit hash algorithm operation of SHA1. The INaud length of the input signal should not be less than 512 bits, so as to avoid reducing the randomness of the output value of 0uthash. In addition, in order to increase the difficulty of guessing the random number sequence, the high-resolution time of the system is read separately.

Only SHA-1 is used for values below the second.

Timehash = SHA—1 (SysTimehigh_res) At this time, the high-resolution time SysTimehigh-res below the system second is subjected to SHA-1 operation to obtain a 160-bit hash value Timehash and then X0R calculation processing with Outhash obtained previously. The XOR operation does not pad the data.

RandomOutput = Outhash XOR Timehash. Therefore, the random number generator uses the thermal noise on the sound card to generate a random output value length of 512 bits. If the random number length is not enough for the application system, the application system only needs to repeat Call] Continued pages (Notes on the use of continuation pages when the invention description page is insufficient) 12 580797

This random number generator can obtain the required random number length. Generally speaking, this output value will have the following random quality indicators:

Entropy value Chi square distribution of random numbers Predicted values Arithmetic mean Monte Carlo Peripheral correlation coefficient is about 98% 25 ～ 75% 97 ～ 99% Accuracy rate 0.5 54.0% Error rate approx. 0.01 The above Entropy value indicates random numbers The theoretical density of information, the closer it is to 100% in theory, the less repeated redundant information, that is, the more ideal the degree of clutter; the predicted value of the random number of the Chi square distribution can be used to distinguish that a random number sequence is predictable. Virtual random number sequence, or real random number output that is difficult to predict. If the calculated Chi square predicted value is greater than 99% or less than 1%, it means that the random number sequence can be determined as a non-random output. Between 95% and 99% or between 1 ° /. And 5%, the random number sequence is not sufficiently random. The ideal Chi square prediction value should be between 10% and 90%. A large number of virtual random number sequences are compared with hardware real random number sequences. The value of Chi square confirms that the predicted value of the random number of the Chi square distribution is suitable as an index to distinguish the virtual random number. The random number sequence produced by the present invention has a chi square prediction value between 25 and 75%, that is, it has random unpredictability. The operation flow for generating random number sequences according to the present invention can be divided into two sequences: thermal noise acquisition and random value processing. There is also the operation of setting the system to read the length of the microphone signal each time. The process is as follows: 1. The setting process of the microphone signal reading length (as shown in Figure 1): Continued page (the description page of the invention is insufficient, please note the female page) 13 580797

1. Perform a loop to record the Entropy value of the microphone noise, read the length from 64 to 256 bytes, and increase 32 bytes each time; 2. Read the microphone signal three times according to the given length, and calculate its average Entropy Value and weighted value, the weighting algorithm is, W = E + 17 / L. Among them, E means the average value of Entropy, and L means the length of the read signal. Record these data in the data structure. 3. Find the 3 most weighted values from all analysis records and write them to the system configuration file in descending order; write The input format is: L = dd, E = d.dd, W = d. (D represents a number)

Second, the thermal noise acquisition process (as shown in Figure 2) ·· 1 · If it is the first time to call, read the system configuration configuration file to obtain the INPUT_BUFFER_SIZE value; 2 · According to the system configuration file Signal input length, read the thermal noise on the microphone inaud = ReadMicrophone (INPUT_BUFFER_SIZE); If this is the first time that the random number generation function is executed, you need to perform 3 · to analyze the microphone hardware status, otherwise skip to 4 .;

3. Calculate the absolute value of the correlation coefficient of the input signal value. If the value is greater than 0.9, it should be regarded as a hardware failure. The system returns a relative error message and terminates the execution; INscc = ComputeCorrelationCoefficient (INAUD);

If (INscc > = 0.9) return HW ERROR;

If (0.7 < INSCC < 0.9) {TRY-NO ++; If (TRY-NO > 5) return HW_ERROR; GOTO step 3;}; if the absolute value of the correlation coefficient is between 0.7 and 0.9 Time, continue to read more sample signals to further determine the state of the hardware; absolute values below 0 · 7 belong to the normal state of the hardware, 0 continued pages (when the invention description page is insufficient, please note and use Continued) 14 580797

Pass through the effects of the sound card ’and mark the status _ mark. If the absolute value of the correlation coefficient obtained by touch sampling is between 0.7 and 0.9, the system will return an error message and terminate the execution after 5 sampling tests; 4. Store the input signal value and complete the thermal noise operation Take the program. 3. Random number post-processing (as shown in Figure 3): 1. SHA-1 512bit hash function value for calculating real thermal noise data:

Outhash = SHA_512 (INaud);

2. Read the high-resolution time of the system, and only take the value below the second for 8 ^ _ 丨 calculation; the high-resolution time below this second can reach _sec〇nd on Unix and the system can also reach millisecond Accuracy;

Timehash = SHA—l (SysTimehigh_res); 3. Use the hash value of step 1. and the hash value of the high-resolution time obtained in step 2. as x〇r operation, and do not pad the data. ;

RandomOutput = Outhash XOR Timehash;

4. Clear the input signal value and return the result obtained in step 3. as the output value; complete the random post-processing procedure. A preferred example of the present invention: The random number generator of the present invention can be applied to the handshake negotiation and key production procedure of an SSL server to meet the requirements on the speed and quality of the random number generator. The procedure is as follows: SSL server Continued page (If the description page is insufficient, please note and use the continuation page) 15 580797

Step 1: Receive the Hello message from the client; ClientHello ◄- Step 2: Because this is the first time that the SSL server performs the random number function of the sound card, you need to read the system configuration settings file to obtain the INPUTJBUFFEI ^ SIZE parameter value;

Step 3: The system calls the thermal noise capture function of the sound card random number generator INaud = ReadMicrophone (INPUT_BUFFER_SIZE); Step 4: Because this is the first time that the sound card random number function is performed, the correlation coefficient of the signal needs to be calculated Absolute value to determine the absolute value of the hardware state correlation coefficient;

INscc = ComputeCorrelationCoefficient (INAUD);

If (INscc > = 0.9) return HW_ERROR;

If (0.7 < INscc < 0.9) {TRY—NO ++;

If (TRY NO > 5) return HW ^ ERROR; GOTO step three;} Set the hardware inspection flag; Continued page (If the description page of the invention is insufficient, please note and use the continued page) 580797

Step 5: Calculate the SHA-l 512-bit hash function value of the thermal noise data. Outhash = SHA-512 (INaud); Step 6: Read the high-resolution time below the system second and perform the operation to obtain a hash of 16 shirts. value

Timehash = SHA—1 (SysTimehigh res), Step 7: Perform the XOR operation on the hash value of the signal in step 4 and the hash value of the high-resolution time obtained in step 5, and do not perform data length padding processing here;

RandomOutput = Outhash XOR Timehash; Step 8:

ServerHello.random = RandomOutput ί Then the SSL server composes a Hello message in response with the random value obtained in step 6 and returns it to the client. SSL server

ServerHello

ServerHelloDone □ Continued page (If the description page is insufficient, please note and use the continued page) 17

The random number generator of the sound effect card provided by the present invention has the following advantages when compared with other manufacturing technologies. The random number generator of the sound effect card of the present invention is generally used on a personal computer or the workstation. The installed sound effect card, the noise value on the microphone of Na is used as the source of the ship number sequence, there is no need to install a specific driver or make any settings, which greatly improves the convenience of use and still has no fetters Hard-aged number f, in addition, in the treatment of _ Yincheng's sexuality ', we adopt the knot design obtained after theoretical and experimental analysis to process Wei, taking into account the performance of the system under the best random number quality. 2. The present invention uses a random number generator of a sound card, which uses unpredictable thermal noise as the original value of the random number, and then uses a carefully designed cryptographic hash algorithm to hash the signal. The real random number obtained is Sequences have excellent performance on various major randomness evaluation indicators. 2. The present invention uses a random number generator of a sound card to perform a post-processing operation of a random number sequence using a 512-bit one-way hash algorithm, which not only improves the quality index value of the random number, but also eliminates the reverse cracking of the original Noise sequence possible. In addition, the high-resolution time value and random number output value of the system are read for mask processing, which further increases the difficulty of guessing the random number sequence. The above detailed description is a specific description of a feasible embodiment of the present invention, but this embodiment is not intended to limit the patent scope of the present invention. Any equivalent implementation or change that does not depart from the technical spirit of the present invention should be included in Within the scope of the patent in this case. In summary, this case is not only technically innovative, but also enhances the above-mentioned multiple effects over conventional items. 'It should have fully met the requirements for novel and progressive statutory invention patents. Apply according to law' and ask your office to approve this. Application for invention patents, to encourage invention, to the utmost. [] Continued pages (Please note and use continuation pages when the invention description page is insufficient.)

[Brief description of the drawings] With reference to the following detailed description of the preferred embodiment of the present invention and the accompanying drawings, the technical content of the present invention and its purpose and effect can be further understood; the drawings of the embodiments are: Figure- FIG. 1 is a flowchart of setting a microphone signal reading length of a random number generator using a sound effect card according to the present invention; FIG. 1 * is a flowchart of thermal noise manipulation and judgment of the random number generation using the effect card; The third is a flowchart of processing the random number generator using the sound card after the random number sequence.

Claims (1)

580797 patent application scope 1. A sound card number generator, a method for generating high-quality random number sequences on the sound card, its architecture includes-sound cards and microphones commonly installed on personal computers and Unk workstations Signal acquisition interface, the library of the random number generator and a tool program to determine the length of the signal read by the microphone; the process of producing random numbers can be divided into the pre-processing thermal noise acquisition and analysis and random number sequence Two parts are processed afterwards. 2. The random number generator using a sound card as described in item 1 of the scope of patent application, wherein the pre-processing procedure includes a method for reading the length of the thermal noise and a method for judging the usable state of the microphone hardware. The read length of the noise must be set to achieve the best random number quality under the performance required by the system; the method of determining the hardware availability status is based on the correlation between the signals before and after, which can avoid the problem caused by the sound card hardware. Malfunction caused poor quality of random numbers. 3. The random number generator using a sound effect card as described in item 1 of the scope of the patent application, wherein the post-processing program includes the selection of a 512-bit one-way hash function, the random number sequence combined with the high-resolution time of the system, and other processing Steps: Selecting a longer hash value is based on the g and theoretical and experimental analysis results, which can increase the entropy of the random number sequence; while combining the high-resolution time of the operating system to form the final random number sequence, the random number can be increased in proportion. The difficulty of guessing a sequence of numbers. 4. The random number generator using a sound card as described in item 1 of the scope of patent application, wherein the tool program for setting the microphone to read the signal length, and the signal analysis and evaluation process and settings can be found in a short reading Take the parameter that produces a larger entropy at length. 20