A one-way import method for external files based on offline media fingerprint binding

Abstract

This invention provides a one-way import method for external files based on offline media fingerprint binding, belonging to the field of network security and data exchange technology. The method includes: S1. Multi-dimensional physical feature signal acquisition; S2. Fingerprint entropy calculation; S3. Initialization of a chaotic mapping system; S4. Construction of a sparse hash tree integrity structure; S5. Cross-validation fusion calculation and update; S6. Block-by-block integrity verification; S7. One-way file data writing operation; S8. Generation of a final import report. This invention ensures that there is no traceable information channel between the external media and the internal system after the import is completed.

Description

Technical Field

[0001] This invention relates to the fields of network security and data exchange technology, and in particular to a one-way import method for external files based on offline media fingerprint binding. Background Technology

[0002] In sectors with extremely high information security requirements, such as finance, critical internal networks are typically strictly physically isolated from external public networks or other networks with lower security levels to prevent external network attacks and data leaks. However, these high-security internal networks inevitably need to import data from external sources, such as receiving software updates, intelligence information, and transaction data. Currently, a common data import method involves using offline storage media such as Universal Serial Bus flash drives, Secure Digital Storage Cards, and optical discs as transfer tools. External data is manually copied to the media, which is then connected to a terminal on the internal network for further data transfer.

[0003] This traditional method has several inherent risks: First, the offline storage medium itself may be implanted with malicious hardware or firmware, which can launch attacks when connected to the internal system; second, the data stored on the medium may be tampered with or damaged during the transfer process, and the internal system lacks an effective integrity verification mechanism during import; third, once the data import process is completed, due to software remnants or caching, a traceable information channel may be left between the external medium and the internal system, creating a potential risk of data leakage.

[0004] Existing one-way data import technologies, such as those based on optically isolated one-way transmission cards, while solving the physical channel problem of reverse information leakage, fail to deeply bind the physical identity of the medium to data integrity. This makes them ineffective against combined attacks involving medium forgery and data tampering. Furthermore, existing file integrity verification methods, such as those using fixed hash checksums, are limited in their security strength due to their static and independent nature when facing targeted attacks.

[0005] Therefore, there is an urgent need for a secure import method that can deeply integrate the unique physical identity of offline storage media, the data integrity of the files to be imported, and the strict one-wayness of the import process, in order to achieve a higher level of security for data exchange in physically isolated network environments. Summary of the Invention

[0006] This invention provides a one-way import method for external files based on offline media fingerprint binding. It generates a unique media fingerprint by collecting the inherent physical characteristics of the offline storage medium. This fingerprint is used as the initial seed of a chaotic dynamic system to drive the generation of a dynamic key stream. Simultaneously, the file to be imported is segmented and a sparse algebraic verification tree embedding the dynamic key is constructed. Then, the chaotic key stream and the root hash of the verification tree are cross-fused for verification, achieving deep binding between the media fingerprint, file data, and dynamic key. Finally, under the guarantee of a one-way physical channel, the file data, after block-by-block integrity verification, is imported, and self-verification is performed after import. The entire method strictly adheres to the principle of one-way data flow, ensuring that there is no traceable information channel between the external medium and the internal system after import.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A one-way import method for external files based on offline media fingerprint binding includes:

[0009] S1. Collect physical characteristic signals of the offline storage medium and generate a set of multi-dimensional physical characteristic signals that characterize the unique physical identity of the offline storage medium;

[0010] S2. Construct a medium fingerprint feature vector based on the multidimensional physical feature signal set, and calculate the fingerprint entropy value of the medium fingerprint feature vector;

[0011] S3. Initialize the chaotic system based on the medium fingerprint feature vector and the fingerprint entropy value, and perform iterative calculations to generate a chaotic state vector sequence and a chaotic key stream matrix;

[0012] S4. Divide the external file to be imported into several data blocks to form a set of file data blocks, construct a sparse hash tree using the chaotic key stream matrix, and obtain the hash value of the root node of the sparse hash tree.

[0013] S5. Perform cross-validation fusion calculation using the root node hash value, the chaotic state vector sequence, and the chaotic key stream matrix to generate a fusion verification vector and cross-validation result, and perform path update on the sparse hash tree to obtain the sparse hash tree with the path update completed and the new root node hash value.

[0014] S6. When the cross-confirmation result is passed, the file data block set is subjected to block-by-block integrity verification using the fusion verification vector, and a block index list of verified blocks and a block index list of verified blocks are generated.

[0015] S7. Based on the verified block index list, retrieve the corresponding data block from the file data block set and write it to the internal storage system through the physical layer unidirectional transmission channel. At the same time, set the write operation status flag bit according to the verified failed block index list and generate a written data block record table.

[0016] S8. When the write operation status flag meets the preset conditions, the written data blocks are read back from the internal storage system according to the written data block record table, and self-verification is performed using the fusion verification vector, the new root node hash value, and the sparse hash tree that has completed the path update, and a final import report is generated.

[0017] In this specification, the specific method for acquiring the physical characteristic signals of the offline storage medium in S1 is as follows: a scanning excitation signal is applied to the offline storage medium and a response signal is acquired. From the response signal, the zero-crossing rate characteristic value and peak factor characteristic value output by the time domain analysis channel, the spatial variance output by the spatial domain analysis channel, and the amplitude attenuation at nine frequency points output by the frequency domain analysis channel are extracted. The zero-crossing rate characteristic value, the peak factor characteristic value, the spatial variance, and the nine amplitude attenuation values ​​together constitute the multidimensional physical characteristic signal set. The spatial variance is obtained by acquiring the phase values ​​of the response signals at multiple equally spaced locations in the horizontal, vertical, and normal directions of the plane of the offline storage medium, and calculating the spatial variance of all phase values.

[0018] In this specification, the specific method for constructing the medium fingerprint feature vector based on the multidimensional physical feature signal set in S2 is as follows: First, the multidimensional physical feature signal set is arranged into an original feature row vector. Then, a standardized mapping process is performed on each component of the original feature row vector, which is subtracted from the global mean and divided by the global standard deviation, to obtain a standardized feature row vector. Next, a nonlinear enhancement transformation based on the hyperbolic tangent function is performed on each component of the standardized feature row vector to obtain an enhanced feature row vector. The enhanced feature row vector is then assigned as the medium fingerprint feature vector. The fingerprint entropy value is calculated as follows: the sum of squares of all components in the medium fingerprint feature vector is calculated. A normalized probability distribution is constructed based on the proportion of the square of each component in the sum of squares. Then, the Shannon entropy value is calculated using the normalized probability distribution as the fingerprint entropy value.

[0019] In this specification, the chaotic system is a four-dimensional hyperchaotic system, whose state equation contains four system parameters. The numerical combination of these system parameters causes the four-dimensional hyperchaotic system to be in a hyperchaotic state with at least two positive Lyapunov exponents. The specific method for initializing the chaotic system based on the medium fingerprint feature vector and the fingerprint entropy value is as follows: the first four components are extracted from the medium fingerprint feature vector as the initial state variables of the chaotic system, and the fingerprint entropy value is introduced as a perturbation factor into the fourth initial state variable. The iterative operation uses the fourth-order Runge-Kutta method to perform discrete iterations, discarding the transient state points generated in the first few iterations and retaining the state points generated in subsequent iterations as the chaotic state vector sequence. The specific method for generating the chaotic keystream matrix is ​​as follows: a composite operation including division, modulo, and floor is performed on the four state components of each state vector in the chaotic state vector sequence, mapping each state vector to a byte value, and all byte values ​​constitute the chaotic keystream matrix.

[0020] In this specification, the specific method for constructing a sparse hash tree using the chaotic keystream matrix is ​​as follows: For each data block in the file data block set, a hash calculation is performed using the data block and the key byte value dynamically extracted from the chaotic keystream matrix as input to obtain the hash value of the leaf node. For empty leaf nodes that do not contain actual data blocks, a preset fixed hash value is assigned. Starting from the leaf node layer, the hash calculation is performed layer by layer upwards for each parent node using the hash values ​​of its left and right child nodes and the key byte value dynamically extracted from the chaotic keystream matrix as input, until the hash value of the root node is calculated. The method for dynamically extracting the key byte value from the chaotic keystream matrix is ​​to take the modulo of the current node's level number and position number with the number of rows and columns of the chaotic keystream matrix, respectively.

[0021] In this specification, the specific method of cross-validation fusion calculation is as follows: The root node hash value is expanded into a byte sequence; each byte of the byte sequence is XORed with the corresponding key byte extracted from the chaotic keystream matrix to obtain a forward fusion value; the forward fusion value is used to perform a sinusoidal weighted superposition of the state components in the chaotic state vector sequence; each byte of the forward fusion value is XORed with another set of key bytes extracted from the chaotic keystream matrix to obtain a feedback fusion value; the root node hash value and the feedback fusion value are concatenated and then hashed to obtain the new root node hash value; the forward fusion value is then XORed with... The feedback fusion value combination is the fusion verification vector; the specific method of performing path update on the sparse hash tree is as follows: for all parent nodes on the path from the root node to each non-empty leaf node in the sparse hash tree, the hash values ​​of their left and right child nodes, the key byte values ​​dynamically extracted from the chaotic key stream matrix, and the feedback fusion value are used as inputs to re-perform hash calculation, and update the hash values ​​of each node layer by layer from bottom to top; the method of generating the cross-confirmation result is as follows: calculate the Hamming distance between the new root node hash value and the forward fusion value, compare the Hamming distance with a preset threshold, and if it does not exceed the preset threshold, the cross-confirmation result is passed.

[0022] In this specification, the specific method for block-by-block integrity verification is as follows: For each data block in the file data block set, calculate the concatenated hash value of the data block and the corresponding key byte value dynamically extracted from the chaotic key stream matrix to obtain the current verification hash value; concatenate the current verification hash value with the forward fusion value in the fusion verification vector and perform a hash operation to obtain the merged verification value; concatenate the hash value of the leaf node corresponding to the data block in the sparse hash tree with the forward fusion value and perform a hash operation to obtain the baseline merged value; compare the merged verification value with the baseline merged value; if they match, the data block is determined to have passed verification; if they do not match, the data block is determined to have failed verification; add the index number of the verified data block to the verified block index list, and add the index number of the failed data block to the failed block index list.

[0023] In this specification, the physical layer unidirectional transmission channel is implemented by an optical isolator or a diode unidirectional conduction circuit. The optical isolator uses the photoelectric conversion principle to convert electrical signals into optical signals for unidirectional propagation. The diode unidirectional conduction circuit uses the unidirectional conductivity of a semiconductor PN junction to block reverse electrical signals. In step S7, during the process of writing the verified data blocks into the internal storage system through the physical layer unidirectional transmission channel, the index number of each data block, as well as the write start timestamp and write end timestamp, are recorded. The index numbers, write start timestamps, and write end timestamps of all written data blocks together constitute the written data block record table. When the verification failure block index list is not empty, the write operation status flag is set to a partial failure state. When the verification failure block index list is empty, the write operation status flag is set to a full success state. When a hardware error occurs during the writing process, the write operation status flag is set to a write error state.

[0024] In this specification, the write operation status flag meeting the preset condition means that the write operation status flag is in a fully successful state or a partially failed state; the specific method of self-verification is as follows: for each data block in the written data block record table, the data block is read back from the internal storage system to obtain the readback block data, and the hash value of the concatenation of the readback block data and the corresponding key byte value dynamically extracted from the chaotic key stream matrix is ​​calculated to obtain the readback verification hash value; the readback verification hash value is concatenated with the feedback fusion value in the fusion verification vector and a hash operation is performed to obtain the confirmation hash value; the updated leaf node hash value corresponding to the data block is obtained from the sparse hash tree that has completed the path update as the theoretical leaf hash value; the confirmation hash value and the theoretical leaf hash value are compared. If they are consistent, the data block is marked as having passed self-verification; if they are inconsistent, the data block is marked as having failed self-verification; the number of data blocks that have passed self-verification and the number of data blocks that have failed self-verification are counted, and the integrity self-verification success rate is calculated.

[0025] In this specification, the final import report includes the write operation status flag, the written data block record table, the self-verification status flag for each written data block, the number of data blocks that passed self-verification, the number of data blocks that failed self-verification, the integrity self-verification success rate, and the start and end timestamps of the import process.

[0026] In summary, the present invention has at least the following beneficial effects:

[0027] At the media identity authentication level, this invention collects the inherent physical response characteristics of offline storage media in multi-dimensional physical space (time domain, spatial domain, frequency domain) to construct a media fingerprint feature vector with uniqueness and non-cloning. This elevates the security trust root of the imported data from a logical identifier (such as a serial number or digital certificate) that can be tampered with or forged to a non-replicable feature at the physical layer, fundamentally eliminating attacks carried out through media forgery.

[0028] At the dynamic key derivation level, this invention uses the medium fingerprint feature vector as the initial state variable of the four-dimensional hyperchaotic system and the fingerprint entropy value as the perturbation factor. By taking advantage of the extreme sensitivity of the hyperchaotic system to the initial conditions, the tiny differences in the physical fingerprint are amplified exponentially in the phase space to generate a chaotic key stream matrix with high randomness and long-term unpredictability. This makes any physical tampering or replacement of the medium cause an avalanche change in the entire key stream, thus preventing attackers from replacing the medium or data without changing the key stream.

[0029] At the level of file integrity protection, this invention embeds the dynamic elements in the chaotic key stream matrix as hash salt values ​​into the hash calculation process of each node of the sparse hash tree. This makes the integrity verification value of the data block no longer a static and unchanging fixed value, but a binding value that changes dynamically with the medium fingerprint, effectively resisting pre-computation collision attacks and length expansion attacks targeting fixed hash functions.

[0030] At the level of deep binding of multiple elements, this invention injects the information feedback of the root node hash value into the evolution trajectory of the chaotic state vector through forward fusion computation, embeds the information of the chaotic key stream into the update process of the root node hash value through reverse feedback computation, and keeps the new root node hash value consistent with the tree structure through path update of the sparse hash tree. A two-way closed-loop mathematical coupling relationship is established among the physical fingerprint, chaotic key stream and file data. Any independent tampering of any single element will spread to the entire system through the mathematical propagation path and be captured by the detection mechanism, realizing the collaborative defense capability that the static independent verification mechanism in the prior art does not have.

[0031] In terms of security during the import process, this invention achieves unidirectional flow of data signals through a physical layer unidirectional transmission channel (optical isolator or diode unidirectional conduction circuit), which completely cuts off the physical path of reverse information leakage from the electrical principle level and eliminates the hidden dangers of residual channels or concealed backhaul channels that may exist in software layer control.

[0032] In terms of system fault tolerance and reliability, this invention employs a dual verification mechanism of block-by-block integrity verification before writing and self-verification through readback after writing. This mechanism can detect different types of data anomalies, such as transmission errors, physical damage to the medium, and writing hardware failures, at different stages of data import. It can also pinpoint the anomaly to the specific data block, thus preventing the entire import task from failing due to local data corruption and providing system administrators with accurate fault diagnosis criteria. This effectively improves the reliability and maintainability of data import tasks in high-security environments. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of a one-way import method for external files based on offline media fingerprint binding.

[0034] Figure 2 This is a schematic diagram of the process for acquiring multidimensional physical feature signals and constructing media fingerprints.

[0035] Figure 3 This is a schematic diagram of the chaotic key stream generation process.

[0036] Figure 4 This is a schematic diagram of the cross-validation fusion and hash tree path update process. Detailed Implementation

[0037] The technical solution of this invention will be described in detail below. The method of this invention can be executed by an import control host, which is connected to a dedicated card reader / drive interface for reading offline storage media and an internal storage system. The connection to the internal storage system is achieved through a physical layer unidirectional transmission channel. For example... Figure 1 As shown, the specific steps of the one-way import method for external files based on offline media fingerprint binding are as follows.

[0038] S1. Perform multi-dimensional physical characteristic signal acquisition on the offline medium. It should be noted that the input of S1 is the offline storage medium to be operated on. The offline storage medium is a physical carrier with solid-state storage, magnetic storage, or optical storage functions, including but not limited to universal serial bus flash drives, secure digital memory cards, optical discs, magnetic tapes, or disks. The output of S1 is a set of multi-dimensional physical characteristic signals. This set consists of twelve independent signal components in three dimensions: time domain, spatial domain, and frequency domain.

[0039] The specific implementation process of S1 is as follows: First, the offline storage medium is physically connected to the import control host through a dedicated card reader or driver interface. The import control host has a built-in hardware module combining a high-precision analog-to-digital converter and a signal generator. Then, the signal generator applies a scanning excitation signal with a frequency continuously and linearly varying between 100 kHz and 10 MHz to the storage chip or disk surface of the offline storage medium. The voltage amplitude of the scanning excitation signal is fixed at 1.5 volts. Simultaneously, the high-precision analog-to-digital converter synchronously acquires the response signal of the offline storage medium under the action of the scanning excitation signal at a sampling rate of 5 mega-samples per second. The response signal is preprocessed by a fourth-order Butterworth bandpass filter with a passband range set to 80 kHz to 12 MHz to eliminate high-frequency noise from environmental electromagnetic interference and low-frequency drift from the power supply module. The preprocessed response signal is then fed in parallel into three different signal analysis channels, corresponding to time domain analysis, spatial domain analysis, and frequency domain analysis, respectively.

[0040] In the time-domain analysis channel, the zero-crossing rate characteristic value of the preprocessed response signal is measured. and peak factor eigenvalue Among them, the zero-crossing rate eigenvalue The peak value is defined as the number of times the signal waveform crosses the zero amplitude point within a unit time window of 1 millisecond; peak factor characteristic value. The value is defined as the ratio of the absolute peak value of the signal to the root mean square (RMS) value of the signal. In the spatial domain analysis channel, a stepper motor built into the control host is used to drive the offline storage medium relative to the signal generator to generate a three-dimensional spatial micro-displacement, with a step size of 0.1 micrometers. Response signal phase values ​​are collected at ninety-nine equally spaced locations along the transverse, longitudinal, and perpendicular dimensions of the medium plane, and then the spatial variance of these ninety-nine phase values ​​is calculated. As a spatial domain phase shift signal, a Fast Fourier Transform is performed on the preprocessed response signal in the frequency domain analysis channel to calculate the amplitude attenuation at a specific set of frequency points in the spectrum. This specific set of frequency points includes nine frequency points: 100 kHz, 250 kHz, 500 kHz, 750 kHz, 1 MHz, 2.5 MHz, 5 MHz, 7.5 MHz, and 10 MHz. The amplitude attenuation at each frequency point is recorded as the corresponding component value of the frequency domain resonant attenuation signal and sequentially labeled as follows: to The twelve signal components obtained from the three channels are combined into a multidimensional physical feature signal set. The mathematical expression for this set is: .

[0041] In some embodiments, considering that the response signals of offline storage media in different physical regions and at different frequencies may exhibit different signal-to-noise ratios and stability due to media aging, local physical damage, or changes in ambient temperature, this embodiment adds a physical feature signal component quality assessment and dynamic weighted optimization step between step S1 and step S2 to screen out signal components with higher quality and assign them higher weights to participate in the construction of the media fingerprint feature vector, thereby improving the robustness and stability of the fingerprint under media aging or environmental change conditions.

[0042] The specific implementation of this step is as follows: In step S1, the multidimensional physical feature signal set is completed. After the collection, Each of the twelve signal components was individually evaluated for quality. The quality evaluation was based on two dimensions: the time stability coefficient and the signal-to-noise ratio coefficient for each signal component. For Each signal component subscript The set of positive integers whose component indices range from 1 to 12, respectively, corresponds to , , , to First, calculate the time stability coefficient of the signal component. Time stability coefficient The acquisition method is as follows: During the response signal acquisition process in step S1, the same frequency point or the same physical location is measured repeatedly every 0.5 seconds, for a total of ten consecutive measurements, to obtain ten measurement values. Then calculate the standard deviation of these ten measurements. Compared with the average The ratio, that is:

[0043] ;

[0044] in For the first The standard deviation of ten repeated measurements of a signal component. For the first The arithmetic mean of ten repeated measurements of a signal component. The aforementioned positive minimum constant has a value of Its function is to prevent the denominator from being zero. Time stability coefficient. The value of is between 0 and 1. The larger the value, the smaller the fluctuation of the signal component over time and the higher the stability.

[0045] Then calculate the first... Signal-to-noise ratio coefficient of each signal component Signal-to-noise ratio coefficient The acquisition method is as follows: before the signal generator applies the scanning excitation signal in step S1, a 1-millisecond environmental noise signal is acquired through a high-precision analog-to-digital converter, and the root mean square value of the environmental noise signal is calculated. Then, after applying the scanning excitation signal, the first... The root mean square value of each signal component The environmental noise signal and the response signal are acquired under the same environmental conditions, with the time interval between the two acquisitions not exceeding 2 seconds to ensure the relative stability of the noise level. Signal-to-noise ratio coefficient. The calculation formula is:

[0046] ;

[0047] in For the first The root mean square value of each signal component This represents the root mean square value of the environmental noise signal. This is the aforementioned positive minimum constant. Signal-to-noise ratio coefficient. The value range is also between 0 and 1. The larger the value, the less the signal component is affected by environmental noise and the higher the signal quality.

[0048] Time stability coefficient and signal-to-noise ratio coefficient The first is obtained through multiplicative fusion. Overall quality score of each signal component :

[0049] ;

[0050] in For the first The overall quality score of a signal component is a real number ranging from 0 to 1. The closer the value is to 1, the higher the quality of the signal component.

[0051] After calculating the overall quality score of all twelve signal components Next, set a quality acceptance threshold. Its value is 0.6. The overall quality score... Greater than or equal to The signal components are marked as high-quality components, and the overall quality score is calculated. Less than The signal components are marked as low-quality components. Instead of being discarded directly, the weights of these low-quality components in the subsequent normalization mapping and nonlinear enhancement transformation in step S2 are attenuated.

[0052] Specifically, before the normalization mapping process in step S2, for Each signal component Multiplied by a dynamic weighting factor The weighted signal component values ​​are obtained. :

[0053] ;

[0054] in For the dynamically weighted first Each signal component value For the first The dynamic weighting coefficients of each signal component are calculated as follows:

[0055] ;

[0056] in This represents the maximum value among all twelve composite quality scores. This is the aforementioned positive minimum constant; This is a temporary index variable, taking values ​​from 1 to 12, used to represent the position of a component in the overall quality score sequence. Dynamic weighting coefficients. The value of ranges from 0 to 1, with the highest quality signal component having a weighting coefficient of 1, and the weighting coefficient of the lower quality signal component being closer to 0. After dynamic weighting, the weighted signal component values ​​are... This constitutes a dynamically weighted set of multidimensional physical feature signals. :

[0057] ;

[0058] Then with Replace the original As input to step S2, it is used to construct the medium fingerprint feature vector and calculate the fingerprint entropy value.

[0059] This embodiment introduces a signal quality assessment and dynamic weighting mechanism, so that the construction of the medium fingerprint feature vector no longer treats all physical feature components equally, but assigns differentiated weights to each component based on its actual signal quality. This mechanism allows the stability and identifiability of the medium fingerprint to be maintained even if the response signal quality of some frequency points or physical areas of the offline storage medium deteriorates due to long-term use, physical aging, or changes in ambient temperature. This is achieved through weighted compensation of other high-quality components, significantly improving the robustness and reliability of the solution in real-world engineering environments. Furthermore, this quality assessment mechanism itself utilizes repeated measurement statistics and signal-to-noise ratio analysis of the signal over time.

[0060] S2. Construct a medium fingerprint feature vector based on the set of physical feature signals and calculate the fingerprint entropy. It should be noted that the output of S2 is the medium fingerprint feature vector. fingerprint entropy Among them, the medium fingerprint feature vector It is a mathematically compact expression of the unique physical identity of the offline storage medium, namely the fingerprint entropy value. It is a quantitative indicator that measures the richness of the fingerprint feature vector information of the medium. For example... Figure 2 As shown, the specific implementation process of S2 is divided into the medium fingerprint feature vector construction stage and the fingerprint entropy calculation stage.

[0061] In the phase of constructing the medium fingerprint feature vector, the first step is to start from a set of multidimensional physical feature signals. Extract all twelve signal component values ​​and arrange them into an original feature row vector. :

[0062] ;

[0063] Then, the original feature row vectors are processed. Each component in the algorithm undergoes a normalization mapping process. The purpose of this process is to unify signal components with different physical dimensions to the same numerical scale, thereby eliminating the impact of dimensional differences on subsequent calculations. The formula for the normalization mapping process is:

[0064] ;

[0065] Among them, subscript This is a component index, and its value range is a set of positive integers from 1 to 12; Represents the original feature row vector The first in One component; The first term obtained after normalization mapping is represented as... One standard component value; This indicates the statistical results obtained from a pre-established offline media fingerprint feature database. The global mean of each component; This indicates the number of fingerprints obtained from statistics in the offline media fingerprint feature database. The global standard deviation of each component; It is a small positive constant used to prevent the denominator from being zero, and its value is set to . After standardization mapping, all standard component values Constructing standardized feature row vectors :

[0066] ;

[0067] Next, the standardized feature row vectors are processed. The purpose of performing a nonlinear enhancement transform is to amplify the subtle differences in fingerprint features between different offline storage media, thereby improving the discriminative power of the media fingerprints. The nonlinear enhancement transform uses a variant of the hyperbolic tangent function:

[0068] ;

[0069] in, The first one obtained after nonlinear enhancement transformation One enhanced component value; It is the hyperbolic tangent function; For the first The gain coefficients corresponding to each component; For the first The bias coefficients corresponding to each component. The gain coefficients With the bias coefficient All of these parameters were pre-calculated using statistical distribution parameters from the offline media fingerprint feature database. The specific calculation method is as follows: , After nonlinear enhancement transformation, all enhancement component values Constructing enhanced feature row vectors :

[0070] ;

[0071] Finally, the feature row vectors will be enhanced. Directly assigning the value to the medium fingerprint feature vector ,Right now:

[0072] ;

[0073] In the fingerprint entropy calculation stage, the medium fingerprint feature vector is used. The information richness of the medium fingerprint is quantified by the values ​​of its various components. First, the feature vector of the medium fingerprint is calculated. The sum of squares of all twelve components:

[0074] ;

[0075] in, Represents the feature vector of medium fingerprint The first in One component; This is the sum of squares of the twelve components. Then, the normalized probability distribution is calculated. :

[0076] ;in, Indicates the first The proportion of each component of energy in the total energy satisfies The probability normalization condition is then determined. Finally, the fingerprint entropy value is calculated. Its calculation formula uses the continuous form of Shannon entropy:

[0077] ;

[0078] in, Represents the logarithmic function with base 2; The aforementioned positive minimum constant has a value of Its function is to prevent in When the logarithmic function takes the value of zero, it tends towards negative infinity. (Fingerprint entropy value) The value range is from 0 to Between these ranges, the larger the value, the more significant the medium fingerprint feature vector. The more uniform the information distribution, the stronger the uniqueness of the medium fingerprint.

[0079] S3. Initialize the chaotic mapping system using the medium fingerprint feature vector and drive the generation of the chaotic key stream. It should be noted that the output of S3 is a sequence of chaotic state vectors. With chaotic key stream matrix S3 employs a dynamic key derivation algorithm based on a four-dimensional hyperchaotic system. This algorithm, characterized by its extreme sensitivity to initial conditions, long-term unpredictability, and high-dimensional dynamic complexity, is suitable for generating dynamic key materials with high security strength. For example... Figure 3 As shown, the specific implementation process of S3 is divided into the chaotic system initialization stage, the chaotic state iteration stage, and the key stream generation stage.

[0080] In the initialization phase of a chaotic system, the first step is to analyze the medium fingerprint feature vector. Extract the first four component values , , , As the four initial state variables of the chaotic system, and the fingerprint entropy value These serve as the initial control parameters for the chaotic system. Specifically, the state equation of a four-dimensional hyperchaotic system is defined as:

[0081] ;

[0082] in, There are four state variables. These represent the derivatives of the four state variables with respect to time. The first system parameter is set to 36; This is the second system parameter, and its value is set to 3; This is a third system parameter, and its value is set to 28. The fourth set of system parameters is set to 1. This set of classical parameters enables the system to be in a hyperchaotic state, possessing at least two positive Lyapunov exponents, ensuring extreme complexity of dynamic behavior and extreme sensitivity to initial conditions, thus providing a high-dimensional key space and an unpredictable long-term evolution trajectory. The discretized iterative solution of the above state equations adopts the fourth-order Runge-Kutta method, with a discrete iteration step size of... Set to 0.001 seconds. To generate a sufficient number of keystreams, the total number of iterations is [number missing]. The number of iterations is set to 17,000. The four initial state variables are assigned values ​​as follows:

[0083] ;

[0084] Among them, fingerprint entropy value Introduced into the fourth initial state variable As a perturbation factor, multiplied by a coefficient This is to adjust the influence of entropy to a numerical level comparable to that of the state variables, thereby ensuring that the fingerprint entropy has a detectable impact on chaotic dynamics.

[0085] In the chaotic state iteration phase, starting from the initialized state variables, 17,000 iterations are continuously executed according to the fourth-order Runge-Kutta method discrete iteration scheme, resulting in a discrete state sequence of length 17,001. To eliminate the transient effects of the chaotic system's transition state, the state points generated in the first 1,000 iterations are discarded, and only the 16,000 state points from the 1,001st to the 17,000th iteration are retained as the effective chaotic state vector sequence. The effective chaotic state vector sequence Each element in the vector is a four-dimensional state vector, i.e. , where subscript A superscript represents a set of positive integers whose position indices in a sequence range from 1 to 16000; This represents the vector transpose operation.

[0086] During the key stream generation phase, for a valid chaotic state vector sequence Each four-dimensional state vector The four state components are mapped to a key byte value through a nonlinear fusion function, thereby generating a chaotic key stream sequence. The expression of the nonlinear fusion function is:

[0087] ;

[0088] in, In the chaotic key stream sequence, the first... One key byte value; This indicates a modulo operation with 256, ensuring the result falls within the byte range of 0 to 255; This indicates the floor function; To prevent extremely small positive constants with a denominator of zero, after the above nonlinear fusion mapping, all key byte values... A one-dimensional chaotic keystream sequence of length 16000 bytes is constructed. This one-dimensional chaotic keystream sequence is then rearranged into a matrix with 16 bytes per row, forming a matrix of size... Chaotic key stream matrix Among them, the matrix row index The value range is the set of positive integers from 0 to 999, and the matrix column index is... The value of is a set of positive integers from 0 to 15. The in the matrix is ​​the . Line number Column elements Equal to the first in the chaotic key stream sequence One key byte value.

[0089] S4. Perform chunking on the target external file and construct a sparse hash tree integrity structure. It should be noted that the input to S4 is the target external file itself and the chaotic keystream matrix output from S3. The output of S4 is a set of file data blocks. Hash value of the root node of the sparse hash tree S4 employs a method combining a chaotic keystream matrix. The sparse hash tree construction algorithm, which serves as a source of dynamic hash parameters, differs from traditional complete binary hash trees in that it only stores the hash node paths corresponding to non-empty data blocks, thereby significantly reducing storage overhead without sacrificing verification capabilities. The specific implementation process of S4 is divided into a file segmentation stage and a sparse hash tree construction stage.

[0090] During the file chunking phase, the complete binary data stream of the target external file is first read and represented as a sequence of bytes. ,in, Represents the first in the file data stream byte; subscript The value ranges from 0 to Integers; This indicates the total length of the file in bytes. Then, it is divided into fixed block sizes. For byte sequences Perform non-overlapping segmentation, the fixed block size The value is set to 4096 bytes. The total number of blocks obtained after splitting is:

[0091] ;

[0092] in, This indicates the rounding up operation; This represents the total number of blocks. When the last block's data length is insufficient... When needed, padding is performed using bytes with a value of 0 to make up the total. Length. After being divided into blocks, each block is assigned an integer index number starting from 0. ,in The value range is from 0 to A set of integers. All blocks constitute the file data block set. ,in, Indicates that the index number is The byte array corresponding to each block, the length of each block byte array is 1. .

[0093] During the sparse hash tree construction phase, a tree of height is constructed. A sparse hash tree, the height of which The calculation formula is:

[0094] ;

[0095] in, This represents the base-2 logarithmic function. Rounding up and adding 1 ensures the tree height is sufficient to accommodate all possible block indices. The total number of leaf nodes in a sparse hash tree is... Each leaf node corresponds to a possible index number position, but only those index numbers are considered. Less than The leaf nodes contain the actual data blocks. For each leaf node containing the actual data block... The leaf node, its leaf hash value The calculation method is as follows:

[0096] ;

[0097] in, This represents the 256-bit version of the secure hash algorithm, which outputs a 256-bit (32-byte) hash value. This indicates a byte sequence concatenation operation; Represents the chaotic key stream matrix Take the first one from the middle Line number The key byte value of the column participates in the hash calculation as a dynamic hash salt value. This represents the modulo operation. For empty leaf nodes that do not contain actual data blocks, their leaf hash value is fixed to a pre-agreed constant value. The constant value The 32-byte constant sequence is represented in hexadecimal as: 0x0 ...

[0098] After calculating the hash values ​​of all leaf nodes, the hash values ​​of parent nodes are calculated layer by layer upwards from the leaf node level until the hash value of the root node is calculated. For any parent node in the tree, the hash value of its left child node is denoted as... Its right child node hash value is denoted as Then the hash value of the parent node The calculation method is as follows:

[0099] ;

[0100] in, The parent node is numbered at its level in the tree, starting from level 0 for leaf nodes and increasing upwards level by level. Number the parent node from left to right in the current level, starting from 0 and incrementing; Indicates based on hierarchy and location From the chaotic key stream matrix The corresponding key byte is dynamically selected as the hash salt value. After calculating layer by layer upwards, the hash value of the root node of the sparse hash tree is finally obtained. The root node hash value Essentially, it refers to the entire target external file data content passing through a chaotic key stream matrix. A compact digital summary of integrity after dynamic disturbances.

[0101] S5. Perform cross-validation fusion calculation between the chaotic keystream matrix and the hash value of the sparse hash tree root node, and update the hash tree path. It should be noted that the output of S5 is the fused verification vector. Cross-confirmation flag And the updated hash value of the new root node The core value of S5 lies in establishing a bidirectional interactive verification mechanism between the chaotic keystream matrix and the hash value of the root node of the sparse hash tree. Through mathematical fusion calculations, it ensures that the two are logically bound together, preventing the incorrect combination and use of chaotic keystream matrices and file data from different sources or batches. For example... Figure 4 As shown, the specific implementation process of S5 is divided into the forward fusion calculation stage, the reverse feedback calculation stage, the cross-confirmation and determination stage, and the sparse hash tree path update stage.

[0102] In the forward fusion computation phase, the hash value of the root node of the sparse hash tree is first... Expanding the sequence of bytes into a 32-byte sequence, each byte is represented as... subscript The value range is the set of integers from 0 to 31. Then, from the chaotic keystream matrix... Extract the first 32 key bytes, i.e. to These key byte values ​​are marked sequentially as ,in Also using integer indices from 0 to 31. Then, a forward fusion XOR operation is performed:

[0103] ;

[0104] in, This represents the bitwise XOR operator; For the first Thirty-two forward-fused bytes. Concatenate into a forward merge value of length 32 bytes Then use the forward fusion value. For the first component of each state vector in the chaotic state vector sequence Perform a weighted update:

[0105] ;

[0106] in, This is the value of the first state component after forward weighting update; The function is a sine function. This weighted update injects the root node hash value into the state evolution of the chaotic system, causing the subsequent chaotic trajectory to be influenced by both the medium fingerprint and the file data content. After the weighted update is completed, the first state component value of the t-th state vector in the chaotic state vector sequence is... Replace with the updated version This ensures that the first state component value in the subsequent chaotic state vector sequence is calculated using the updated value.

[0107] In the reverse feedback calculation phase, the forward fusion value is used. Again with the chaotic keystream matrix The last 32 key bytes are subjected to a feedback XOR operation. Specifically, this is performed from the chaotic keystream matrix. Extract the last 32 key bytes located at row indices 47 to 48 and column indices 0 to 15, and mark them as... ,in Integer indices from 0 to 31. Perform a feedback XOR operation:

[0108] ;

[0109] in, Represents the forward fusion value The first in 1 byte; For the first One feedback fusion byte. Thirty-two feedback fusion bytes. Concatenate them into a feedback fusion value with a length of 32 bytes. Then use the feedback fusion value. hash value of the root node of a sparse hash tree The update is performed, resulting in a new root node hash value:

[0110] ;

[0111] in, This is the new root node hash value after feedback updates. This update operation embeds the information of the chaotic key stream matrix in reverse into the root hash of the sparse hash tree, so that the root hash value also carries the information of the chaotic dynamic state.

[0112] During the sparse hash tree path update phase, in order to create a new root... To ensure correct use for subsequent integrity self-verification, the sparse hash tree constructed in S4 must undergo path updates. The update process involves: identifying the paths from the root node to every non-empty leaf node in the S4 sparse hash tree; for each parent node on these paths, ... It participates in its hash recalculation as a new additional salt value. Specifically, for the level of Location is The updated hash value of the parent node on the path. Defined as:

[0113] ;

[0114] This update calculation is performed layer by layer from bottom to top until the result is obtained. A consistent new root is established. For all nodes along these paths, their old hash values ​​are replaced with the updated hash values. This step ensures the structural integrity of the entire sparse hash tree is maintained. It has been re-verified.

[0115] During the cross-confirmation phase, the updated hash value of the new root node will be fed back. With forward fusion value Perform eventual consistency verification. First, calculate the consistency metric. :

[0116] ;

[0117] in, Represents two bytes and The Hamming distance between two pairs of data is the number of bits whose corresponding bit values ​​differ. (Consistency metric) The theoretical value range is an integer between 0 and 256. Then, a consistency acceptance threshold is set. The threshold is set to 64. The selection of this threshold is based on the following: For a 256-bit hash value, the expected total Hamming distance between two independent random byte sequences is 128. Setting the acceptance threshold to 64 requires that the total Hamming distance not exceed half of this expected value, which is a relatively strict consistency requirement. Less than or equal to Then the cross-confirmation flag will be set. Set it to logical truth and confirm that cross-validation passes; if Greater than Then the cross-confirmation flag will be set. Set it to a logical false value and determine that cross-validation has failed. Simultaneously, set the forward fusion value... Fusion value with feedback Together they form a fusion verification vector .

[0118] S6. Perform block-by-block integrity verification on the set of file data blocks based on the fused check vector. It should be noted that the output of S6 is a list of blocks that passed verification. and the list of block indexes for verification failures The purpose of S6 is to utilize fused check vectors before writing file data to the internal system. The integrity and correctness of each data block are independently verified block by block to ensure that there is no data corruption due to transmission errors or physical damage to the medium.

[0119] The specific implementation process of S6 is as follows: First, check the cross-confirmation flag. The value, if If the result is a logical false value, all subsequent processing will be terminated immediately, and an overall verification failure will be reported to the system. No further operations will be performed by S6. If the result is logically true, then the block-by-block integrity verification process continues.

[0120] In the block-by-block integrity verification process, from the fused verification vector Extract forward fusion values As a verification benchmark. For file data block sets. Each block in ,in The block index number is a number ranging from 0 to 1. Given a set of integers, perform the following verification calculation process.

[0121] First, from the chaotic key stream matrix Extracting and partitioning indexes Corresponding dynamic hash salt value Then calculate the current verification hash value of the block:

[0122] ;

[0123] in, Indicates that the index number is The verification hash value calculated at the current verification time for each block is 256 bits (32 bytes) long. Then, the current verification hash value... With forward fusion value Perform a merge hash operation to obtain the merged verification value:

[0124] ;

[0125] in, This represents the merged hash result after concatenating the current verification hash value and the forward fusion value; its length is also 256 bits, or 32 bytes. Then, the original leaf hash value is obtained from the leaf node corresponding to this block when constructing the sparse hash tree in S4. The original leaf hash value With the same forward fusion value Perform the same merge hash operation to obtain the baseline merged value:

[0126] ;

[0127] in, This represents the baseline hash result after concatenating the original leaf hash value with the forward fusion value. Finally, a comparison is made. and The two are considered completely equal if they are identical in every single bit of the 256-bit block. The block is then deemed to have passed verification, and its index is assigned. Add to the list of blocks that have passed verification If the two are not completely equal, that is, at least one bit is inconsistent, then the block verification is deemed to have failed, and the block's index number is changed. Add to the list of failed verification block indices After performing the above verification calculation process on each block, a complete list of verified block indices is obtained. List of block indexes with verification failures .

[0128] In some embodiments, an updated sequence of chaotic state vectors is used before the block-by-block integrity verification process begins. The first state component value in the calculation yields a dynamic verification adjustment factor. :

[0129] ;

[0130] in For the updated chaotic state vector sequence, the th The first state component value of a state vector. The total number of data blocks. The dynamic verification adjustment factor. With forward fusion value The bitwise XOR operation yields the dynamically adjusted verification baseline value. :

[0131] ;

[0132] in Indicates will The operation of quantizing to an 8-bit unsigned integer. This is a hash function for the 256-bit version of the secure hash algorithm. Then, using... Replace the original This serves as the verification benchmark value in subsequent block-by-block integrity verification.

[0133] S7. Perform a one-way write operation on the file data based on the block verification result list. It should be noted that the output of S7 is a table of written data block records. With write operation status flags The core objective of S7 is to write all validated data blocks into the internal storage system in a physically unidirectional manner, while strictly prohibiting any reverse flow of data from the internal system to the outside. The specific implementation process of S7 is divided into three sub-stages: the write preparation stage, the data block write stage, and the status recording stage.

[0134] During the write preparation phase, the list of failed block indexes is checked first. Is the list empty? If the list is non-empty, meaning it contains at least one element, it indicates that at least one block has failed integrity verification, and the operation status flag will be written in this case. Set to the enumeration value PARTIAL_FAILURE to indicate partial failure, and continue to execute subsequent write operations on verified blocks to recover data as much as possible; if If the list is empty, meaning it contains no elements, then the operation status flag will be written. Set to the enumeration value ALL_SUCCESS to represent all successes.

[0135] During the data chunking and writing phase, the import control host writes data chunks into the internal storage system via a physical layer unidirectional transmission channel. This physical layer unidirectional transmission channel is implemented using either an optical isolator or a diode unidirectional conduction circuit. The optical isolator utilizes the photoelectric conversion principle to convert electrical signals into optical signals for transmission, and then converts the optical signals back into electrical signals at the receiving end. Because the optical path is unidirectional, it physically cuts off the path for reverse electrical signals. The diode unidirectional conduction circuit utilizes the unidirectional conductivity of the semiconductor PN junction, allowing current to flow only from the transmitting end to the receiving end; any reverse voltage is blocked by the high-impedance state. Both implementations ensure that data signals can only be transmitted from the transmitting end of the import control host to the receiving end of the internal storage system, while any reverse electrical signals are physically blocked, thus meeting the security requirement of strictly unidirectional data flow.

[0136] For the verified block index list Each index number in From file data chunks Extract the corresponding block byte array from and the block byte array Together with its index number Together, the data is written to a predetermined storage area in the internal storage system via the physical layer unidirectional transmission channel. This predetermined storage area is a pre-allocated logical volume or directory path within the internal storage system used to store the imported data; its specific path information is determined by the import control host during the initialization phase based on the system configuration. During the writing process, the index number of each block is recorded after it is written. And write the start timestamp and write end timestamp The write start timestamp and write end timestamp are both provided by a high-precision real-time clock module inside the control host, with a timing accuracy of milliseconds. The index numbers, write start timestamps, and write end timestamps of all written blocks together constitute the written data block record table. Each entry in this record table is in the form of an ordered triple. .

[0137] During the status recording phase, after completing the write operations for all verified blocks, the write operation status flag will be set. The final value is determined as: if the block index list previously failed due to verification. Not empty but will If set to PARTIAL_FAILURE, this flag will remain at PARTIAL_FAILURE; if all blocks pass verification and are successfully written, this flag will remain at ALL_SUCCESS. If any hardware error occurs during the write process, such as hardware failure at the sender or receiver of the unidirectional transmission channel, insufficient storage space in the internal storage system, or a checksum error, causing a block to fail to be written, the write operation status flag will be reset. Forced to be set to the enumeration value WRITE_ERROR, indicating a write error, and recorded in the written data block record table. Only records of blocks that were successfully written are retained; records of blocks that failed to be written are not included. middle.

[0138] In some embodiments, in order to prevent attackers from carrying out replay attacks by intercepting and replaying previously successfully imported data blocks in a physically isolated network environment, this embodiment adds an anti-replay attack step based on timestamp binding and hash chain verification between steps S7 and S8, so that the data blocks of each imported session are mathematically bound to the global timestamp of the current session, and any attempt to replay old data will be detected and rejected due to timestamp mismatch.

[0139] The specific implementation of this step is as follows: Before the data block unidirectional write operation is executed in step S7, a global timestamp for the current session is first generated by the high-precision real-time clock module inside the import control host. The global timestamp An example format is the number of milliseconds that have elapsed since January 1, 2000, 00:00:00 to the present moment, and its value is a 64-bit unsigned integer. This global timestamp It remains unchanged within a single import session, serving as a unique time identifier for that import session.

[0140] In the data block writing stage of step S7, the block index list that has passed verification is... Each index number in In dividing the byte array into blocks Before being written to the predetermined storage area of ​​the internal storage system via the physical layer unidirectional transmission channel, a time-bound hash operation is first performed on the block to generate the time-bound hash value of the block. :

[0141] ;

[0142] in For index number The time-bound hash value of the block, This is a hash function for the 256-bit version of the secure hash algorithm. This is a byte sequence concatenation operator. For index number The byte array corresponding to the block, The global timestamp of the current import session. To obtain from the chaotic key stream matrix The corresponding key byte value is dynamically retrieved from the database. This time-bound hash value. The block data, global timestamp, and chaotic key byte value are mathematically bound together, and any change to any one of them will cause the hash value to change.

[0143] Then, the time of this block is bound to a hash value. As additional metadata for this block, along with the block byte array and index number Together, they are written to a predetermined storage area in the internal storage system via the unidirectional transmission channel of the physical layer. During the writing process, It is stored in the metadata segment corresponding to the block in the predetermined storage area, forming a one-to-one correspondence with the block data itself.

[0144] After all verified blocks have been written, a globally bound hash value is generated. This global binding hash value aggregates the time-binding hash values ​​of all blocks written in this import session:

[0145] ;

[0146] in For the data block record table that has been written The index numbers of all blocks recorded in the record. The total number of data blocks that have been written. This is the global timestamp of the current import session. This is the global binding hash value. Attached to the final import report In this context, it serves as the global integrity anchor point for this import session. Simultaneously, it is stored in a designated metadata area of ​​the internal storage system, along with the written data block record table, for retrieval during the self-verification process. The globally bound hash value... The generation is based on all records written into the data block record and maintains strict consistency with the written data block record.

[0147] In the self-verification process of step S8, the globally bound hash value stored during the write phase is read from the specified metadata area of ​​the internal storage system. This is used for subsequent comparisons. For the data block record table that has already been written... For each block, the block byte array is read back from a predetermined storage area of ​​the internal storage system. At the same time, it also reads back the time-bound hash value stored in the metadata segment corresponding to the block. Then use the block byte array obtained from the readback. Global timestamp of the current import session and chaotic key byte value Recalculate the theoretical time-bound hash value of this block. :

[0148] ;

[0149] Compare and If the two are completely equal, it means that the timestamp binding verification of the block is successful. The block is newly written data in this session, not replay data from the previous session. If the two are not completely equal, it means that the timestamp binding verification of the block is failed, the block is marked as a suspicious replay block, and the self-verification status of the block is marked as "verification failed - timestamp binding abnormal".

[0150] Simultaneously, after the self-verification process is completed, the time-binding hash values ​​and global timestamps of all blocks obtained from the readback are used. Recalculate the global binding hash value :

[0151] ;

[0152] Compare Read from a specified metadata area of ​​the internal storage system If the two are completely equal, the global integrity verification for this import session is confirmed to have passed; if they are not equal, it means that the data has been written to the data block record table. Inconsistencies exist between the records in the database and the actual data in the storage system, potentially indicating data tampering or replay attacks. In such cases, an operation status flag will be written. Forced to be set to the enumeration value INTEGRITY_ALERT, which represents an integrity exception.

[0153] This embodiment embeds a global timestamp into the hash binding calculation of each data block, ensuring that each data block in an import session has a unique and non-reusable time identity. Even if an attacker successfully intercepts all data blocks transmitted through the physical layer unidirectional transmission channel in a previous import session and retransmits them in subsequent sessions, the global timestamp will still prevent data loss. The situation has changed; the time-bound hash value for replaying data blocks has been changed. Will be compared with the recalculated theoretical value Inconsistencies are reliably identified and rejected by the self-verification mechanism. Simultaneously, a globally bound hash value is used. It provides an additional layer of global integrity verification; any tampering or replay of any single chunk will result in... and The mismatch further enhances the security of the scheme against replay attacks. This anti-replay mechanism complements the integrity verification mechanisms of the chaotic keystream matrix and sparse hash tree, providing dual protection in two orthogonal directions: the time dimension and the data dimension.

[0154] S8. Perform post-write block integrity self-verification and generate the final import report. It should be noted that the output of S8 is the final import report. The purpose of S8 is to read back the written data blocks from the internal storage system after the data has been written and recalculate their hash values ​​to match the updated root node hash value in S5. The data is compared with the sparse hash tree that completes the path update in S5 to ultimately confirm the integrity and correctness of the written data.

[0155] The specific implementation process of S8 is as follows: First, check the write operation status flag. The value. If If the value is WRITE_ERROR, the final import report will be generated directly. The report includes error codes and a table of written data blocks. The report includes all entries and a clear error message string. Specific error message strings include, but are not limited to, "Hardware write error - One-way transmission channel failure," "Hardware write error - Insufficient internal storage space," or "Hardware write error - Verification error," clearly indicating the specific reason for the write failure. The entire process ends after the report is generated, and subsequent S8 integrity self-verification operations are not performed.

[0156] If write operation status flag bit If the result is PARTIAL_FAILURE or ALL_SUCCESS, the integrity self-verification process continues. This applies to records already written to the data block table. The index number recorded for each entry in the database. The corresponding block data is read back from a predetermined storage area of ​​the internal storage system, and the resulting block byte array is recorded as follows. Then from the chaotic keystream matrix Extracting from the index number Corresponding dynamic hash salt value Calculate the readback verification hash value of the readback block:

[0157] ;

[0158] in, The index number read back from the internal storage system is The readback verification hash value is calculated from the blocks. Then, the readback verification hash value is... Fusion value with feedback Perform merge hash operation:

[0159] ;

[0160] in, This represents the confirmation hash value obtained by concatenating the readback verification hash value and the feedback fusion value. Then, from the sparse hash tree where the path update has been completed in S5, the index number is retrieved. The updated leaf node hash value corresponding to the block is recorded as the theoretical leaf hash value. ,Right now:

[0161] ;

[0162] in, This indicates that during the sparse hash tree path update phase of S5, after passing through... The index number obtained after updating the additional salt value calculation is: The new hash value of the leaf node. The theoretical leaf hash value. From the updated new root node hash value The data is directly read from the corresponding sparse hash tree, where all internal nodes have been synchronized and updated. Since S5 has already completed the full path update from the leaf to the root, therefore... and They are mathematically consistent.

[0163] Compare and Whether they are completely equal, that is, whether every bit of the two is identical. If they are completely equal, the self-check status of the block is marked as "check passed"; if they are not completely equal, the self-check status of the block is marked as "check failed".

[0164] After completing the self-verification of all written blocks, count the number of blocks that passed the self-verification. and the number of blocks that failed self-checking And calculate the integrity self-verification success rate. The calculation formula is:

[0165] ;

[0166] in, This represents the number of blocks that passed the self-verification. This represents the number of blocks that failed self-checking. The integrity self-verification success rate is expressed as a percentage, and its value ranges from 0% to 100% as a real number.

[0167] Final Import Report The content includes: write operation status flags The final value, already written to the data block record table All entries, the self-verification status flag (i.e., the flag string "verification passed" or "verification failed") for each written block, and the number of self-verification passed blocks. Number of blocks that failed self-check Integrity self-verification success rate And the start and end timestamps of the entire import process. Final import report. It is stored in a designated log area of ​​the internal system in the form of structured text. The structured text format can be JSON, XML, or plain text table format to facilitate post-event auditing and traceability by system administrators.

Claims

1. An external file one-way import method based on offline media fingerprint binding, characterized in that, include: S1. Collect the physical characteristic signals of the offline storage medium and generate a set of multi-dimensional physical characteristic signals that characterize the unique physical identity of the offline storage medium; S2. Construct a medium fingerprint feature vector based on the multidimensional physical feature signal set, and calculate the fingerprint entropy value of the medium fingerprint feature vector; S3. Initialize the chaotic system based on the medium fingerprint feature vector and the fingerprint entropy value, and perform iterative calculations to generate a chaotic state vector sequence and a chaotic key stream matrix; S4. Divide the external file to be imported into several data blocks to form a set of file data blocks, construct a sparse hash tree using the chaotic key stream matrix, and obtain the hash value of the root node of the sparse hash tree. S5. Perform cross-validation fusion calculation using the root node hash value, the chaotic state vector sequence, and the chaotic key stream matrix to generate a fusion verification vector and cross-validation result, and perform path update on the sparse hash tree to obtain the sparse hash tree with the path update completed and the new root node hash value. S6. When the cross-confirmation result is passed, the file data block set is subjected to block-by-block integrity verification using the fusion verification vector, and a block index list of verified blocks and a block index list of verified blocks are generated. S7. Based on the verified block index list, retrieve the corresponding data block from the file data block set and write it to the internal storage system through the physical layer unidirectional transmission channel. At the same time, set the write operation status flag bit according to the verified failed block index list and generate a written data block record table. S8. When the write operation status flag meets the preset conditions, the written data blocks are read back from the internal storage system according to the written data block record table, and self-verification is performed using the fusion verification vector, the new root node hash value, and the sparse hash tree that has completed the path update, and a final import report is generated.

2. The offline media fingerprint bound external file one-way import method of claim 1, wherein, The specific method for acquiring the physical characteristic signals of the offline storage medium in S1 is as follows: a scanning excitation signal is applied to the offline storage medium and a response signal is acquired. The zero-crossing rate characteristic value and peak factor characteristic value output by the time domain analysis channel, the spatial variance output by the spatial domain analysis channel, and the amplitude attenuation at nine frequency points output by the frequency domain analysis channel are extracted from the response signal. The zero-crossing rate characteristic value, the peak factor characteristic value, the spatial variance, and the nine amplitude attenuation values ​​together constitute the multidimensional physical characteristic signal set. The spatial variance is obtained by collecting the response signal phase values ​​of multiple equally spaced locations in the three dimensions of the offline storage medium: horizontal, vertical, and normal directions, and calculating the spatial variance of all phase values.

3. The offline media fingerprint-based external file unidirectional import method of claim 1, wherein, The specific method for constructing the medium fingerprint feature vector based on the multidimensional physical feature signal set described in S2 is as follows: First, the multidimensional physical feature signal set is arranged into an original feature row vector. Then, a standardized mapping process is performed on each component of the original feature row vector, which is subtracted from the global mean and divided by the global standard deviation, to obtain a standardized feature row vector. Next, a nonlinear enhancement transformation based on the hyperbolic tangent function is performed on each component of the standardized feature row vector to obtain an enhanced feature row vector. The enhanced feature row vector is then assigned as the medium fingerprint feature vector. The fingerprint entropy value is calculated as follows: the sum of squares of all components in the medium fingerprint feature vector is calculated. A normalized probability distribution is constructed based on the proportion of the square of each component in the sum of squares. Then, the Shannon entropy value is calculated using the normalized probability distribution as the fingerprint entropy value.

4. The offline media fingerprint-based binding-based one-way import of external files method of claim 1, wherein, The chaotic system is a four-dimensional hyperchaotic system, whose state equation contains four system parameters. The numerical combination of these system parameters causes the four-dimensional hyperchaotic system to be in a hyperchaotic state with at least two positive Lyapunov exponents. The specific method for initializing the chaotic system based on the medium fingerprint feature vector and the fingerprint entropy value is as follows: the first four components are extracted from the medium fingerprint feature vector as the initial state variables of the chaotic system, and the fingerprint entropy value is introduced as a perturbation factor into the fourth initial state variable. The iterative operation uses the fourth-order Runge-Kutta method to perform discrete iterations, discarding the transient state points generated in the first few iterations and retaining the state points generated in subsequent iterations as the chaotic state vector sequence. The specific method for generating the chaotic keystream matrix is ​​as follows: a composite operation including division, modulo, and floor is performed on the four state components of each state vector in the chaotic state vector sequence, mapping each state vector to a byte value, and all byte values ​​constitute the chaotic keystream matrix.

5. The offline media fingerprint-based binding-based one-way import of external files method of claim 1, wherein, The specific method for constructing a sparse hash tree using the chaotic keystream matrix is ​​as follows: For each data block in the file data block set, a hash calculation is performed using the data block and the key byte value dynamically extracted from the chaotic keystream matrix as input to obtain the hash value of the leaf node. For empty leaf nodes that do not contain actual data blocks, a preset fixed hash value is assigned. Starting from the leaf node layer, the hash calculation is performed layer by layer upwards for each parent node using the hash values ​​of its left and right child nodes and the key byte value dynamically extracted from the chaotic keystream matrix as input, until the hash value of the root node is calculated. The method for dynamically extracting the key byte value from the chaotic keystream matrix is ​​to take the modulo of the current node's level number and position number with the number of rows and columns of the chaotic keystream matrix, respectively.

6. The offline media fingerprint bound external file one-way import method of claim 1, wherein, The specific method of the cross-validation fusion calculation is as follows: the root node hash value is expanded into a byte sequence, and each byte of the byte sequence is XORed with the corresponding key byte taken from the chaotic key stream matrix to obtain the forward fusion value; the forward fusion value is used to perform a sinusoidal weighted superposition of the state components in the chaotic state vector sequence. Each byte of the forward fusion value is XORed with another set of key bytes extracted from the chaotic keystream matrix to obtain the feedback fusion value; the root node hash value and the feedback fusion value are concatenated and then hashed to obtain the new root node hash value; the forward fusion value and the feedback fusion value are combined to form the fusion verification vector; the specific method for updating the path of the sparse hash tree is as follows: for all parent nodes on the path from the root node to each non-empty leaf node in the sparse hash tree, the hash values ​​of their left and right child nodes, the key byte values ​​dynamically extracted from the chaotic keystream matrix, and the feedback fusion value are used as inputs to re-perform hash calculation, and the hash values ​​of each node are updated layer by layer from bottom to top; the method for generating the cross-confirmation result is as follows: the Hamming distance between the new root node hash value and the forward fusion value is calculated, and the Hamming distance is compared with a preset threshold. If it does not exceed the preset threshold, the cross-confirmation result is passed.

7. The offline media fingerprint bound external file one-way import method of claim 1, wherein, The specific method of block-by-block integrity verification is as follows: for each data block in the file data block set, calculate the concatenation hash value of the data block and the corresponding key byte value dynamically extracted from the chaotic key stream matrix to obtain the current verification hash value; The current verification hash value is concatenated with the forward fusion value in the fusion verification vector, and a hash operation is performed to obtain the merged verification value. The hash value of the leaf node corresponding to the data block in the sparse hash tree is concatenated with the forward fusion value, and a hash operation is performed to obtain the baseline merged value. The merged verification value is compared with the baseline merged value. If they match, the data block is determined to have passed verification; if they do not match, the data block is determined to have failed verification. The index number of the data block that has passed verification is added to the list of indexes for data blocks that have passed verification, and the index number of the data block that has failed verification is added to the list of indexes for data blocks that have failed verification.

8. The method for one-way import of external files based on offline media fingerprint binding according to claim 1, characterized in that, The physical layer unidirectional transmission channel is implemented by an optical isolator or a diode unidirectional conduction circuit. The optical isolator uses the photoelectric conversion principle to convert electrical signals into optical signals for unidirectional propagation. The diode unidirectional conduction circuit uses the unidirectional conductivity of a semiconductor PN junction to block reverse electrical signals. In S7, during the process of writing the verified data blocks into the internal storage system through the physical layer unidirectional transmission channel, the index number of each data block, as well as the write start time stamp and write end time stamp, are recorded. The index numbers, write start times stamps, and write end times stamps of all written data blocks together constitute the written data block record table. When the verification failure block index list is not empty, the write operation status flag is set to partial failure status; when the verification failure block index list is empty, the write operation status flag is set to full success status; when a hardware error occurs during the write process, the write operation status flag is set to write error status.

9. The method for one-way import of external files based on offline media fingerprint binding according to claim 1, characterized in that, The write operation status flag bit satisfying the preset condition means that the write operation status flag bit is in a fully successful state or a partially failed state; the specific method of self-verification is as follows: for each data block in the written data block record table, read back the data block from the internal storage system to obtain the readback block data, calculate the concatenation hash value of the readback block data and the corresponding key byte value dynamically extracted from the chaotic key stream matrix to obtain the readback verification hash value; The readback verification hash value is concatenated with the feedback fusion value in the fusion verification vector, and then a hash operation is performed to obtain the confirmation hash value. The updated leaf node hash value corresponding to the data block is obtained from the sparse hash tree that has completed the path update as the theoretical leaf hash value; the confirmed hash value is compared with the theoretical leaf hash value. If they match, the data block is marked as having passed self-verification; if they do not match, the data block is marked as having failed self-verification. Count the number of data blocks that pass self-verification and the number of data blocks that fail self-verification, and calculate the integrity self-verification success rate.

10. The method for one-way import of external files based on offline media fingerprint binding according to claim 1, characterized in that, The final import report includes the write operation status flag, the written data block record table, the self-verification status flag for each written data block, the number of data blocks that passed self-verification, the number of data blocks that failed self-verification, the integrity self-verification success rate, and the start and end timestamps of the import process.

Images