A security optimization method for an air gap system

Abstract

The present invention relates to a security optimization method for an air gap system, comprising the following steps: S1, obtaining sensitive data and converting the sensitive data into a string that can be sent; S2, determining the information of the data packet to be transmitted currently based on the leading signal and the load, and controlling the switching power supply to transmit data; S3, if retransmission is required, performing a data retransmission process to retransmit the information transmitted in S2, and if retransmission is not required, updating the sensitive data to be transmitted currently, returning to S2 until all the data packet information is transmitted, and then executing S4; S4, capturing a sound signal and extracting the required information; S5, filtering and correcting the sound signal emitted by the switching power supply; S6, decoding the second spectrum signal, the sensitive data, and optimizing the air gap system. Compared with the prior art, the present invention optimizes the air gap system by simulating the process of a malware handling the sound signal during the operation of a computer switching power supply to obtain the data leakage process of sensitive information.

Description

Technical Field

[0001] The present invention relates to the fields of mobile intelligent perception computing and data security, and in particular to a security optimization method for an air-gap system. Background Art

[0002] In recent years, people's attention to computer malware has been increasing. It is reported that the total number of malware samples registered last year increased by 36%, reaching a record high of 690 million. These malware continuously threaten the information security of the public, enterprises and even governments, bringing great challenges to computer security technology.

[0003] To cope with the threats of these malware and ensure the security of important data from leakage, the most common means is to physically isolate important devices and data from the Internet. Therefore, the concept of an air-gap system has been proposed. Devices in the air-gap system are always isolated from the Internet and devices connected to the Internet to avoid attacks from the Internet. Military networks, such as the Joint Worldwide Intelligence Communication System (JWICS), and the networks of financial organizations, critical infrastructures and commercial industries all belong to the known air-gap systems.

[0004] The research on the attack methods against the air-gap system can further test the security of the air-gap system from various aspects. Only by knowing the possible attack methods can effective prevention be made to prevent possible data leakage. Many studies on the attack against the air-gap system have been proposed. These works use heat, vibration, light or electromagnetic, etc. as side channels to extract data according to the different characteristics of signals when the computer executes different commands. For example, BitWhisper and HOSTPOT use thermal sensors to measure temperature changes; Air-ViBeR uses the accelerometer of a nearby mobile phone to measure the vibration caused by the fan; GSMem, USBee and PowerHammer use magnetic sensors to measure the electromagnetic radiation from the bus bar or power line, etc.

[0005] The above-simulated implicit data leakage methods for the air-gap system all rely on high-sampling-rate sensors as the receiving end to steal data, while the sensors on mobile devices are all low-sampling-rate (≤300 Hz). Since the transmission speed is low (both ≤120 bps) and additional devices are introduced, increasing the difficulty of data export, these methods are not suitable for practical use and it is difficult to optimize the security of the air-gap system. Summary of the Invention

[0006] The objective of the present invention is to provide a security optimization method for an air gap system to overcome the defects of the above-mentioned existing technologies. By simulating the sound signals during the operation of a computer switching power supply by malicious software, the data leakage process of sensitive information is obtained to optimize the air gap system.

[0007] The objective of the present invention can be achieved through the following technical solutions:

[0008] A security optimization method for an air gap system, characterized by including the following steps:

[0009] S1. Obtain sensitive data and convert the sensitive data into a transmissible 0 / 1 string;

[0010] S2. Determine the switching frequency corresponding to the leading signal, use the sensitive data to be transmitted currently as the load, determine the data packet information to be transmitted currently based on the leading signal and the load, and control the switching power supply SMPS to transmit the data packet information to be transmitted currently in the form of sound signals by modulating the CPU. The working state of modulating the CPU is specifically to adjust the CPU switching frequency and the duty cycle of the CPU;

[0011] S3. Based on the inference detection process, determine whether the information transmitted in S2 needs to be retransmitted. If retransmission is required, perform the data retransmission process and retransmit the information transmitted in S2. If retransmission is not required, update the sensitive data to be transmitted currently, return to S2, and continue until all data packet information is transmitted, then execute S4;

[0012] S4. Capture the sound signals emitted by the switching power supply SMPS and extract the required information from the leading signals in the obtained sound signals;

[0013] S5. Based on the first noise spectrum in the required information, filter and correct the sound signals emitted by the switching power supply SMPS. During the filtering and correction process, correction is performed based on the filtered high-frequency and low-frequency signals;

[0014] S6. Decode the second spectral signal after filtering and correction to obtain the required sensitive data, and optimize the air gap system based on the required sensitive data.

[0015] Further, the steps of S1 are specifically as follows:

[0016] S11. Design software and obtain sensitive data based on the software;

[0017] S12. Convert the sensitive data into a transmissible 0 / 1 string, and the encoding and decoding methods during the conversion process are designed according to the requirements of the requester.

[0018] Further, the specific steps of S2 are as follows:

[0019] S21. Determine the lower limit a and upper limit b of the CPU switching frequency according to the current CPU, divide the switching frequency into n intervals, and determine the switching frequency corresponding to the preamble signal based on the divided switching frequencies. The switching frequencies corresponding to the preamble signal are successively: a, a + (b - a) / n, a + (b - a) / (n - 1), ……, b;

[0020] S22. Use the sensitive data that needs to be transmitted currently as the payload. The sensitive data that needs to be transmitted currently is a part of the sensitive data. Combine the payload and the preamble signal to form the packet information that needs to be transmitted currently;

[0021] S23. Modulate the working state of the CPU according to the packet information that needs to be transmitted currently, cause the working state of the switch-mode power supply SMPS to change, generate a sound signal with a specific frequency and specific intensity, carry the packet information that needs to be transmitted currently, and transmit the data;

[0022] Further, the specific steps of S3 are as follows:

[0023] S31. After transmitting a packet of information, determine whether the packet information transmitted in S2 contains interference signals based on the inference detection process. The inference detection process includes monitoring the CPU utilization rate and detecting the historical record of CPU utilization information;

[0024] S32. If it contains interference signals, perform the data retransmission process and retransmit the information transmitted in S2. If retransmission is not required, update the sensitive data that needs to be transmitted currently, return to S2, and continue until all packet information is transmitted, then execute S4;

[0025] Further, the specific steps of S4 are as follows:

[0026] S41. Obtain the sound signal emitted by the switch-mode power supply SMPS based on two microphones. The two microphones obtain two different preamble signals;

[0027] S42. Perform short-time Fourier transform on the two different preamble signals to obtain two first spectral signals, and filter the two first spectral signals to obtain two filtered first spectral signals;

[0028] S43. Divide the first spectral signal before filtering into non-overlapping N first frequency bands, perform spectral subtraction on each first frequency band respectively, and obtain the first noise spectrum of each first frequency band based on the first spectral signal before filtering and the filtered first spectral signal;

[0029] S44. Obtain the first noise of the two spectral signals respectively based on the first noise spectrum, and define the first noise correlation based on the first noise;

[0030] Further, the specific steps of S5 are as follows:

[0031] S51. Perform short-time Fourier transform on the signals of two different loads in the sound signals emitted by the switched-mode power supply (SMPS) obtained by two microphones to obtain two second spectral signals. Divide the second spectral signals into N non-overlapping second frequency bands. Perform spectral subtraction for each second frequency band respectively, and obtain the filtered second spectral signals based on the first noise spectrum and the second spectral signals.

[0032] S52. Define a second noise correlation based on the first noise correlation and the filtered second spectral signals. Determine whether the filtered second spectral signals are noise based on the second noise correlation. If so, update the parameters of the first noise correlation. If not, execute S53.

[0033] S53. Obtain the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal of the filtered second spectral signals.

[0034] S54. Establish the correlation between the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal based on the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal to obtain a mapping relationship. Determine whether the amplitude of the low-frequency sound signal of the filtered second spectral signals meets the condition. If not, obtain the amplitude of the high-frequency sound signal corresponding to the amplitude of the low-frequency sound signal based on the mapping relationship, and correct the filtered second spectral signals based on the amplitude of the high-frequency sound signal.

[0035] Furthermore, the specific steps of S6 are as follows:

[0036] S61. Decode the filtered and corrected second spectral signals to obtain the required sensitive data.

[0037] S62. Optimize the air gap system based on the process of obtaining the required sensitive data in S1 - S6.

[0038] Furthermore, the specific steps of S42 are as follows:

[0039] Perform short-time Fourier transform on two different preamble signals to obtain two first spectral signals. After obtaining the first spectral signals, scan the entire frequency band of the first spectral signals, find all representative frequencies with amplitudes greater than the first threshold in each time frame with a pre-configured length, find subsets A1 and A2 of the representative frequencies corresponding to the two first spectral signals respectively, and filter the two first spectral signals based on subsets A1 and A2 to obtain two filtered first spectral signals.

[0040] Furthermore, during the execution of S2, modulate multiple CPU cores on the linux platform.

[0041] Furthermore, the expression of the correlation between the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal is:

[0042] U = H(f′, Rf, F′(f′, Q))

[0043] Wherein, U is the amplitude of the high-frequency sound signal, H is the mapping relationship between the amplitudes of the high- and low-frequency signals, f′ is the intensity of the low-frequency sound signal, Rf is the frequency resolution, F′ is the inverse function of the mapping relationship between the CPU duty cycle and the corresponding amplitude, and Q is the amplitude of the low-frequency sound signal.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) By modulating the working state of the CPU, the present invention enables the SMPS to generate high-frequency and low-frequency signals simultaneously during operation, realizes filtering and correction of the sound signals emitted by the switched-mode power supply SMPS, better simulates the data correction process during data leakage, and is beneficial to optimizing the air gap system.

[0046] (2) The present invention combines multiple CPU modulation methods, including switching frequency and duty cycle, simulates the means of accelerating data transmission during data leakage, is more in line with the actual situation, and is beneficial to optimizing the air gap system. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a flowchart of the present invention;

[0048] Figure 2 is a data transmission diagram of the present invention;

[0049] Figure 3 is a schematic diagram of the algorithm flow of the CPU modulation method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and gives the detailed implementation manner and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0051] The present invention provides a security optimization method for an air gap system. The flowchart of the method is as Figure 1 shown. The data transmission diagram of the present invention is as Figure 2 shown. The data transmission diagram method includes the following steps:

[0052] S1. Obtain sensitive data and convert the sensitive data into a 0 / 1 string that can be sent.

[0053] S2. Determine the switching frequency corresponding to the preamble signal, use the sensitive data to be transmitted currently as the load, determine the packet information to be transmitted currently based on the preamble signal and the load, and control the switched-mode power supply SMPS to transmit the packet information to be transmitted currently in the form of a sound signal by modulating the CPU.

[0054] S3. Determine whether the information transmitted in S2 needs to be retransmitted based on the inference detection process. If retransmission is required, perform the data retransmission process to retransmit the information transmitted in S2. If retransmission is not required, update the sensitive data that needs to be transmitted currently, return to S2, and continue until all packet information is transmitted, then execute S4.

[0055] S4. Capture the sound signal emitted by the switch-mode power supply (SMPS), and extract the required information from the leading signal in the obtained sound signal.

[0056] S5. Filter and correct the sound signal emitted by the switch-mode power supply (SMPS) based on the first noise spectrum in the required information.

[0057] S6. Decode the second spectral signal after filtering and correction to obtain the required sensitive data, and optimize the air gap system based on the required sensitive data.

[0058] The steps of S1 are specifically as follows:

[0059] S11. Design software for a specific target, and obtain sensitive data based on the software.

[0060] Sensitive data includes, for example, user passwords, user property information, etc. The software is installed in the sending-end machine. The means of obtaining sensitive data include technical means such as intercepting input streams, and different designs are carried out for different scenarios.

[0061] S12. Convert the sensitive data into a sendable 0 / 1 string, and the encoding and decoding methods in the conversion process are designed according to the requirements of the requester.

[0062] The specific steps of S2 are as follows:

[0063] S21. According to the current CPU, determine the lower limit a and upper limit b of the CPU switching frequency, divide the switching frequency into n intervals, and determine the switching frequency corresponding to the leading signal based on the divided switching frequencies. The switching frequencies corresponding to the leading signal are successively: a, a + (b - a) / n, a + (b - a) / (n - 1), ……, b.

[0064] S22. Use the sensitive data that needs to be transmitted currently as the payload. The sensitive data that needs to be transmitted currently is a part of the sensitive data. Combine the payload and the leading signal to form the packet information that needs to be transmitted currently.

[0065] The payload can be 100 bit.

[0066] S23. Modulate the working state of the CPU according to the data packet information to be transmitted currently, causing the working state of the switch-mode power supply (SMPS) to change, generating a sound signal with a specific frequency and specific intensity, carrying the data packet information to be transmitted currently, and transmitting the data. The detailed transmission process is as Figure 3 shown.

[0067] Specifically, modulating the working state of the CPU means adjusting the CPU switching frequency and the duty cycle of the CPU. Adjusting the CPU switching frequency can make the SMPS emit a sound signal with a specific frequency. Specifically, the switching frequency f c of the CPU will affect the change frequency of the current, thereby affecting the low-frequency sound signal f s emitted by the SMPS. And through experiments, it is known that f c = f s . In order to be able to adjust the CPU switching frequency to f c , we can bind a specific thread to a specific CPU, so that the CPU can complete a full-load to sleep state switch within a time of 1 / f c . Specifically, taking the linux platform as an example, each thread can be bound to a specific CPU core through the sched_setaffinity() function. In each switching cycle, the transmitter will control its worker to use the pthread_barrier_wait() function to complete the switch between the CPU full load and the sleep state.

[0068] When adjusting the duty cycle of the CPU, for each individual CPU core, in a switching cycle (1 / f c ), there are two states, namely the full-load working state and the sleep state. Assuming the full-load working duration is t H , and the sleep duration is t L , then it can be known that t H +t L = 1 / f c ; Let θ satisfy which is called the duty cycle. Obviously, the duty cycle should be between 0 and 1. It can be known that changing the duty cycle will not change the size of the switching cycle, so it will not change the sound frequency generated by the SMPS. However, experiments show that the duty cycle will affect the sound intensity of the corresponding frequency, so it can be involved in information transmission.

[0069] During the execution of S2, multiple CPU cores can be modulated on the Linux platform, and specific threads can be bound to different cores to generate composite sound signals with various frequency intensities while transmitting information. Specifically, on the Linux platform, multiple cores can be modulated simultaneously through createAnotherTransmitter(), which can achieve a higher data transmission rate.

[0070] The specific steps of S3 are as follows:

[0071] S31. After transmitting the information of a data packet, based on the inference detection process, determine whether the data packet information transmitted in S2 contains interference signals. The inference detection process includes monitoring the CPU utilization rate and detecting the historical record of CPU utilization information;

[0072] S32. If interference signals are included, perform the data retransmission process to retransmit the information transmitted in S2. If retransmission is not required, update the sensitive data to be transmitted currently, return to S2, and continue until all data packet information is transmitted, then execute S4.

[0073] The specific steps of S4 are as follows:

[0074] S41. Obtain the sound signals emitted by the switch-mode power supply (SMPS) based on two microphones. The two microphones obtain two different preamble signals. The frequency sequence of the preamble signal with a length of n + 1 should be [a, a + c, a + 2c, …, b]. Here, a is the lower limit of the CPU switching frequency, b is the upper limit of the CPU switching frequency, c represents the frequency interval, and the theoretical value is c = (b - a) / n, where n is the number of frequency divisions. The preamble signals received by the two different microphones are represented by X1 p and X2 p respectively.

[0075] S42. Perform short-time Fourier transform on the two different preamble signals to obtain two first spectral signals, and filter the two first spectral signals to obtain two filtered first spectral signals.

[0076] The specific steps of S42 are as follows:

[0077] Perform short-time Fourier transform on the two different preamble signals to obtain two first spectral signals. After obtaining the first spectral signals, scan the entire frequency band of the first spectral signals, find all representative frequencies with amplitudes greater than the first threshold in each time frame with a pre-configured length, find the subsets A1 and A2 of the representative frequencies corresponding to the two first spectral signals from the representative frequencies, and filter the two first spectral signals based on the subsets A1 and A2 to obtain two filtered first spectral signals.

[0078] For X1 pand X2 p Perform a short-time Fourier transform on X1 to obtain the first spectral signal X1 p′ and X2 p′ , and then scan the entire frequency band. In each time frame of d ms, find all the representative frequencies with amplitudes greater than e dB, where e is related to the receiver. Then find a subset A of frequencies that satisfies A = [f1, f2, …, fl], where f i satisfies ||f i - f i-1 | - c| < R f , R f represents the frequency resolution and l = (b - a) / c. It can be seen that after this step, we have obtained A1 and A2 calculated from the signals received by the two microphones respectively. That is to say, we have filtered X1 p′ and X2 p′ to obtain two clean signals D1 and D2 as references for later use.

[0079] S43. Divide the first spectral signal before filtering into N non-overlapping first frequency bands. For each first frequency band, perform spectral subtraction respectively to obtain the first noise spectrum of each first frequency band based on the first spectral signal before filtering and the first spectral signal after filtering.

[0080] Divide X1 p′ and X2 p′ into N non-overlapping frequency bands. For each frequency band, perform spectral subtraction respectively. Therefore, the first noise spectrum of the i-th frequency band can be expressed as where is the estimated noise, b i and e i are the start and end frequencies of the i-th frequency band, α i is the over-subtraction factor of the i-th frequency band, and σ i is the adjustment factor, which can be set separately for each frequency band to customize the noise removal performance.

[0081] S44. Obtain the first noise of the two spectral signals respectively based on the first noise spectrum, and define the first noise correlation based on the first noise.

[0082] Assume that in step S43, the noise estimates obtained from X1 p′ and X2 p′ are and respectively. Then define as the first noise correlation, which contains the characteristics of the noise and will be used in the subsequent denoising process. Each time the leading signal is sent, λ will be updated to ensure the real-time nature of the noise characteristics.

[0083] The specific steps of S5 are as follows:

[0084] S51. Perform short-time Fourier transform on signals of two different loads in the sound signals emitted by the switching power supply SMPS obtained by the two microphones to obtain two second spectrum signals, divide the second spectrum signals into N non-overlapping second frequency bands, perform spectral subtraction on each second frequency band, and obtain a filtered second spectrum signal based on the first noise spectrum and the second spectrum signal.

[0085] This step starts to process the real communication signal. Assume that the signal containing data information received by the two microphones is X1. c and X2 c . Perform a short-time Fourier transform to obtain X1 c′ and X2 c′ , divide it into N non-overlapping second frequency bands, and perform spectral subtraction on each second frequency band independently, according to the formula It can be concluded that is the estimated clean signal, b i 、e i , α i , σ i The settings are the same as those in the definition of S43. After this step, a clean signal is obtained. and That is the second spectrum signal after filtering.

[0086] S52, defining a second noise correlation based on the first noise correlation and the filtered second spectrum signal, and determining whether the filtered second spectrum signal is noise based on the second noise correlation; if so, updating the parameters of the first noise correlation; if not, executing S53.

[0087] Define the second noise correlation δ i =|λ i -λ1 i |, where If δ i >p, we can assume that the current signal is actually noise, such as human conversation or movement, and then i =α i +q to α i The p and q mentioned here will be adjusted according to different receivers.

[0088] S53: Obtain the low-frequency sound signal amplitude and the high-frequency sound signal amplitude of the filtered second spectrum signal.

[0089] Assume that there are n CPU cores and the CPU duty cycle satisfies [Θ1,Θ2,…,Θ n ], the switching frequency corresponding to each core is [f1,f2,…,f n, the intensity of the emitted low-frequency sound signal corresponds to [f′1, f′2, …, f′ n , and their amplitudes are Q i , where Rf is the frequency resolution, then it can be obtained that the intensity of the low-frequency sound signal satisfies f′ i = f i ± RfQ i = F(f i , Θ i ), where F represents the learned mapping relationship between the CPU duty cycle and the corresponding amplitude. Similarly, for the part of the generated high-frequency signal, the amplitude U j of its corresponding frequency satisfies U j = G(f1, f2, …, f n , Θ1, Θ2, …, Θ n ). In sequence, the amplitude of the low-frequency sound signal is Q i , and the amplitude of the high-frequency sound signal is U j .

[0090] S54. Establish the correlation between the amplitudes of the high-frequency and low-frequency signals based on the amplitudes of the low-frequency sound signal and the high-frequency sound signal, obtain the mapping relationship, and determine whether the amplitude of the low-frequency sound signal in the filtered second spectral signal meets the conditions. If not, obtain the amplitude of the high-frequency sound signal corresponding to the amplitude of the low-frequency sound signal based on the mapping relationship, and correct the filtered second spectral signal based on this amplitude of the high-frequency sound signal.

[0091] Integrate the formula for the intensity of the low-frequency sound signal and the formula for the amplitude of the corresponding frequency in S53 to establish the correlation between the amplitudes of the high-frequency and low-frequency signals. The expression of the correlation is:

[0092] U = H(f′, Rf, F′(f′, Q))

[0093] where U is the amplitude of the high-frequency sound signal, H is the mapping relationship between the amplitudes of the high-frequency and low-frequency signals, f′ is the intensity of the low-frequency sound signal, Rf is the frequency resolution, F′ is the inverse function of the mapping relationship between the CPU duty cycle and the corresponding amplitude, and Q is the amplitude of the low-frequency sound signal.

[0094] Integrating the operations of S53 and S54, only by using the preamble signal to establish the mapping relationship H, for the subsequent communication signal with information-carrying payload, only need to check whether its low-frequency intensity meets the conditions to know whether the information is in error. If there is an error, use the high-frequency signal to correct it.

[0095] The specific steps of S6 are as follows:

[0096] S61. Decode the filtered and corrected second spectral signal to obtain the required sensitive data;

[0097] S62. Optimize the air gap system based on the process of obtaining the required sensitive data from S1 - S6.

[0098] The present invention has the advantages of being non - invasive, transparent to users, and using off - the - shelf microphones to receive data, etc., and can better simulate data leakage. The present invention also implements a system on a mobile device, which can achieve a data transmission rate of 2400 bps within one meter and has a bit error rate of only 0.01. Compared with the existing mobile - device - based air gap system, the present invention has a 20 - fold performance advantage.

[0099] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field of the present invention through logical analysis, reasoning, or limited experiments based on the concept of the present invention on the basis of the existing technology should fall within the protection scope determined by the claims.

Claims

1. A security optimization method for an air gap system, characterized in that, It includes the following steps: S1. Obtain sensitive data and convert the sensitive data into a transmissible 0 / 1 string; S2. Determine the switching frequency corresponding to the preamble signal, use the sensitive data to be transmitted currently as the payload, determine the packet information to be transmitted currently based on the preamble signal and the payload, and control the switching mode power supply SMPS to transmit the packet information to be transmitted currently in the form of a sound signal by modulating the CPU. The working state of the modulating CPU is specifically to adjust the CPU switching frequency and the duty cycle of the CPU; S3. Based on the inference detection process, judge whether the information transmitted in S2 needs to be retransmitted. If retransmission is required, perform the data retransmission process to retransmit the information transmitted in S2. If retransmission is not required, update the sensitive data to be transmitted currently, return to S2 until all packet information is transmitted, and then execute S4; S4. Capture the sound signal emitted by the switching mode power supply SMPS and extract the required information from the preamble signal in the obtained sound signal; S5. Based on the first noise spectrum in the required information, filter and correct the sound signal emitted by the switching mode power supply SMPS. During the filtering and correction process, correction is performed based on the filtered high and low frequency signals; S6. Decode the second spectral signal after filtering and correction to obtain the required sensitive data, and optimize the air gap system based on the required sensitive data.

2. The security optimization method for an air gap system according to claim 1, wherein The specific steps of S1 are as follows: S11. Design software and obtain sensitive data based on the software; S12. Convert the sensitive data into a transmissible 0 / 1 string, and the encoding and decoding methods during the conversion are designed according to the requirements of the demander.

3. The security optimization method for an air gap system according to claim 1, wherein The specific steps of S2 are as follows: S21. According to the current CPU, judge the lower limit a and the upper limit b of the CPU switching frequency, divide the switching frequency into n intervals, and determine the switching frequency corresponding to the preamble signal based on the divided switching frequency. The switching frequencies corresponding to the preamble signal are in turn: a, a+(b - a) / n, a+(b - a) / (n - 1), ……, b; S22. Use the sensitive data to be transmitted currently as the payload. The sensitive data to be transmitted currently is a part of the sensitive data, and form the packet information to be transmitted currently by combining the payload and the preamble signal; S23. According to the packet information to be transmitted currently, modulate the working state of the CPU to cause the working state of the switching mode power supply SMPS to change, generate a sound signal with a specific frequency and a specific intensity, carry the packet information to be transmitted currently, and transmit the data.

4. A method for optimizing the security of an air gap system according to claim 3, characterized in that, The specific steps of S3 are as follows: S31. After transmitting a packet of information, based on the inference detection process, judge whether the packet information transmitted in S2 contains interference signals. The inference detection process includes monitoring the CPU utilization rate and detecting the historical record of CPU utilization information; S32. If it contains interference signals, perform the data retransmission process to retransmit the information transmitted in S2. If retransmission is not required, update the sensitive data to be transmitted currently, return to S2 until all packet information is transmitted, and then execute S4.

5. The security optimization method for an air gap system according to claim 3, wherein The specific steps of S4 are as follows: S41. Obtain the sound signals emitted by the switch-mode power supply (SMPS) based on two microphones, and the two microphones obtain two different leading signals; S42. Perform short-time Fourier transform on the two different leading signals to obtain two first spectral signals, filter the two first spectral signals, and obtain two filtered first spectral signals; S43. Divide the first spectral signals before filtering into N non-overlapping first frequency bands, perform spectral subtraction on each first frequency band respectively, and obtain the first noise spectra of each first frequency band based on the first spectral signals before filtering and the filtered first spectral signals; S44. Obtain the first noise of the two spectral signals respectively based on the first noise spectra, and define the first noise correlation based on the first noise.

6. The security optimization method for an air gap system according to claim 5, wherein, The specific steps of S5 are as follows: S51. Perform short-time Fourier transform on the signals of two different loads in the sound signals emitted by the switch-mode power supply (SMPS) obtained by the two microphones to obtain two second spectral signals, divide the second spectral signals into N non-overlapping second frequency bands, perform spectral subtraction on each second frequency band respectively, and obtain the filtered second spectral signals based on the first noise spectra and the second spectral signals; S52. Define the second noise correlation based on the first noise correlation and the filtered second spectral signals, and determine whether the filtered second spectral signals are noise based on the second noise correlation. If so, update the parameters of the first noise correlation; if not, execute S53; S53. Obtain the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal of the filtered second spectral signals; S54. Establish the correlation between the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal based on the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal to obtain a mapping relationship, and determine whether the amplitude of the low-frequency sound signal of the filtered second spectral signals meets the condition. If not, obtain the amplitude of the high-frequency sound signal corresponding to the amplitude of the low-frequency sound signal based on the mapping relationship, and correct the filtered second spectral signals based on the amplitude of the high-frequency sound signal.

7. A security optimization method for an air gap system according to claim 1, characterized in that, The specific steps of S6 are as follows: S61. Decode the filtered and corrected second spectral signals to obtain the required sensitive data; S62. Optimize the air gap system based on the process of obtaining the required sensitive data in S1 - S6.

8. The security optimization method for an air gap system according to claim 5, wherein The specific steps of S42 are as follows: Perform short-time Fourier transform on the two different leading signals to obtain two first spectral signals. After obtaining the first spectral signals, scan the entire frequency band of the first spectral signals, find all representative frequencies with amplitudes greater than the first threshold in each time frame with a pre-configured length, find the subsets A1 and A2 of the representative frequencies corresponding to the two first spectral signals respectively from the representative frequencies, and filter the two first spectral signals based on the subsets A1 and A2 to obtain two filtered first spectral signals.

9. A security optimization method for an air gap system according to claim 1, characterized in that During the execution of S2, modulate multiple CPU cores on the linux platform.

10. A security optimization method for an air gap system according to claim 1, characterized in that, The expression of the correlation between the amplitude of the low-frequency sound signal and the amplitude of the high-frequency sound signal is: U = H(f′, Rf, F′(f′, Q)) Among them, U is the amplitude of the high-frequency sound signal, H is the mapping relationship between the amplitudes of the high- and low-frequency signals, f′ is the intensity of the low-frequency sound signal, Rf is the frequency resolution, F′ is the inverse function of the mapping relationship between the CPU duty cycle and the corresponding amplitude, and Q is the amplitude of the low-frequency sound signal.

Images