Time-domain audio steganography method based on generalized audio intrinsic energy and slight amplitude suppression modification

Abstract

The present invention discloses a time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification. The method comprises the following steps: obtaining carrier audio, segmenting and pre-processing the carrier audio, and cutting the carrier audio into audio segments of a set length; slicing the secret information bits to be embedded into equal parts according to the number of audio segments to obtain secret information bit segments; constructing a steganalysis distortion cost function based on generalized audio intrinsic energy based on a complexity-first criterion, and then proposing a distortion cost design criterion for micro-amplitude suppression modification, combining the two to obtain a final steganalysis distortion cost function; calculating the steganalysis distortion cost of each audio segment, and embedding the secret information bit segment into each audio segment based on a stego-coding method to obtain multiple secret audio segments; and concatenating all the secret audio segments to obtain the final secret audio. The present invention has the advantages of higher steganographic security and better secret audio quality.

Description

Technical Field

[0001] The present invention relates to the technical field of time-domain audio steganography, and in particular to a time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification. Background Art

[0002] With the rapid development of computer and electronic technology in recent years, high-quality lossless audio (time-domain audio) in WAV format has been widely disseminated and shared online. Using this as a medium, audio steganography in temporal domain (ASTD) technology involves embedding secret information into time-domain audio without arousing suspicion from the monitoring party. Currently, mainstream and practical time-domain audio steganography schemes are based on the minimum distortion embedding framework, which primarily consists of two parts: stego-coding and distortion cost function design. With the introduction of efficient and practical stego-coding schemes such as STC and SPC, current research within this framework has focused on the design of stego-distortion cost functions.

[0003] Within this framework, some researchers have proposed a time-domain audio steganography method called DFR (Derivative Filtering Residual Distortion Cost). According to public reports, DFR is currently the most secure time-domain audio steganography algorithm. Specifically, under the guidance of the existing complexity-first criterion, DFR introduces the derivative filter residual to characterize the complexity of the carrier, thereby constructing a corresponding stego-distortion cost function. However, the stego-security performance of this method is often highly dependent on the filter used to obtain the residual. Different filters can result in significant differences in the resulting stego-security performance, making the method unstable. Secondly, although the large amplitude-first (LAF) criterion in the DFR method can prevent the modification of small-amplitude samples, it conflicts to some extent with the complexity-first criterion of content-adaptive steganography. This results in less modification of high-complexity, low-amplitude (non-small-amplitude) samples that are originally suitable for embedding, while more modification of low-complexity, high-amplitude samples that are originally unsuitable for embedding occurs, potentially leading to a sharp decline in stego-security performance. Summary of the Invention

[0004] In order to overcome the defects and shortcomings of the existing time-domain audio steganography algorithm DFR in its stegographic distortion cost function design, the present invention provides a time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification. The intrinsic energy of the carrier is used to characterize the carrier complexity. After applying the micro-amplitude suppression criterion to the complexity priority criterion, the small amplitude sampling points are set as "wet points". This is committed to accurately preventing the small amplitude sampling points from being modified, while not affecting the embedded modification of other amplitude sampling points. This makes the encrypted audio obtained by the present invention more difficult to be detected by the steganalyzer, thereby improving the steganographic security of the present invention.

[0005] The second object of the present invention is to provide a time-domain audio steganography system based on generalized audio intrinsic energy and micro-amplitude suppression modification;

[0006] A third object of the present invention is to provide a computer-readable storage medium;

[0007] A fourth object of the present invention is to provide a computer device.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] The present invention provides a time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification, comprising the following steps:

[0010] Obtain carrier audio, perform segmentation preprocessing on the carrier audio, and cut the carrier audio into audio segments of a set length;

[0011] Slicing the secret information bits to be embedded into equal parts according to the number of audio segments to obtain secret information bit segments;

[0012] Based on the complexity priority criterion, a steganalysis distortion cost function based on generalized audio intrinsic energy is constructed. The steganalysis distortion cost function based on generalized audio intrinsic energy is combined with the distortion cost design criterion of micro-amplitude suppression modification to obtain the final steganalysis distortion cost function.

[0013] The steganalysis distortion cost of each audio segment is calculated, and the STC steganographic coding method based on the minimum distortion steganography framework is used to embed the secret information bit segment into each audio segment, thereby obtaining multiple encrypted audio segments.

[0014] All the encrypted audio clips are spliced ​​together to obtain the final encrypted audio.

[0015] As a preferred technical solution, a steganalysis distortion cost function based on generalized audio intrinsic energy is constructed, specifically including:

[0016] Obtaining a local carrier audio signal, and obtaining different carrier audio frequency components based on spectrum transformation;

[0017] Different weights are set for different carrier audio frequency components, and the weights are used to set exponential adjustment coefficients to calculate the intrinsic energy of the local carrier audio signal for steganography;

[0018] The intrinsic energy of the local carrier audio signal used for steganography is used as the complexity index of the center point of the local carrier audio signal. The steganographic distortion cost of the carrier audio is inversely proportional to the complexity. A steganographic distortion cost function based on the generalized audio intrinsic energy is constructed.

[0019] As a preferred technical solution, the local carrier audio signal is obtained by setting a sliding window to intercept or using mirror filling to obtain the local carrier audio signal.

[0020] As a preferred technical solution, the spectrum transformation adopts any one of discrete cosine transform, discrete Fourier transform or discrete wavelet transform.

[0021] As an optimal technical solution, a steganalysis distortion cost function based on generalized audio intrinsic energy is constructed, which is specifically expressed as:

[0022]

[0023]

[0024] Among them, ρ c represents the distortion cost of the center point of the local carrier audio signal, and λ represents the distortion cost used to prevent ρ c The equilibrium constant is too small, ε represents the stability constant, n represents the number of different carrier audio frequency components, f i represents the amplitude of the i-th carrier audio frequency component, w i represents the weight corresponding to the i-th carrier audio frequency component, p represents the exponential adjustment coefficient, Represents the local carrier audio signal x l Intrinsic energy used for steganography; when calculating the intrinsic steganographic energy of audio, the influence of the DC frequency component is excluded, that is, f1 is discarded.

[0025] As a preferred technical solution, the index adjustment coefficient is expressed as:

[0026]

[0027]

[0028] Among them, psehf represents the energy percentage of high-frequency components in each carrier, k represents the selection threshold of high-frequency components, k≥2, is the average amplitude of all carrier audio frequency components in each carrier.

[0029] As a preferred technical solution, the steganographic distortion cost function based on generalized audio intrinsic energy is combined with the distortion cost design criterion of micro-amplitude suppression modification to obtain the final steganographic distortion cost function, which is specifically expressed as:

[0030]

[0031] Among them, |x c | is the center point x of the local carrier audio signal c The amplitude of the sample point is suppressed, T is the amplitude threshold of the modified sampling point, ψ is the distortion cost of the wet point under the micro-amplitude suppression criterion, w i represents the weight corresponding to the i-th carrier audio frequency component, p represents the exponential adjustment coefficient, n represents the number of different carrier audio frequency components, ε represents the stability constant, ρ c Represents the distortion cost of the center point of the local carrier audio signal.

[0032] In order to achieve the above second purpose, the present invention adopts the following technical solutions:

[0033] The present invention provides a time-domain audio steganography system based on generalized audio intrinsic energy and micro-amplitude suppression modification, comprising: a carrier audio acquisition module, a carrier audio preprocessing module, a secret information bit preprocessing module, a steganalysis distortion cost function construction module, a steganalysis distortion cost calculation module, a steganalysis encoding module, and an audio splicing module;

[0034] The carrier audio acquisition module is used to acquire carrier audio;

[0035] The carrier audio preprocessing module is used to perform segmentation preprocessing on the carrier audio, cutting the carrier audio into audio segments of a set length;

[0036] The secret information bit preprocessing module is used to equally slice the secret information bits to be embedded according to the number of audio segments to obtain secret information bit segments;

[0037] The steganalysis distortion cost function construction module is used to construct a steganalysis distortion cost function based on generalized audio intrinsic energy based on a complexity priority criterion, and combine the steganalysis distortion cost function based on generalized audio intrinsic energy with the distortion cost design criterion of micro-amplitude suppression modification to obtain a final steganalysis distortion cost function;

[0038] The steganalysis distortion cost calculation module is used to calculate the steganalysis distortion cost of each audio clip;

[0039] The stego-coding module is used to embed a secret information bit segment into each audio segment based on the STC stego-coding method of the minimum distortion steganographic framework to obtain multiple secret audio segments;

[0040] The audio splicing module is used to splice all the encrypted audio segments to obtain the final encrypted audio.

[0041] In order to achieve the third purpose above, the present invention adopts the following technical solutions:

[0042] A computer-readable storage medium stores a program, which, when executed by a processor, implements the above-mentioned time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification.

[0043] In order to achieve the fourth purpose above, the present invention adopts the following technical solutions:

[0044] A computer device includes a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, it implements the time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification as described above.

[0045] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0046] (1) The existing DFR method uses the derivative filter residual to characterize the complexity of the carrier and construct its steganalysis distortion cost function. The steganalysis security performance of this method usually depends to a large extent on the selection of efficient filters. The present invention abandons the idea of ​​using the carrier signal residual to characterize the carrier complexity, and uses the intrinsic energy of the carrier to characterize the carrier complexity. Based on this, a new steganalysis distortion cost function is designed. This saves the trouble of carefully selecting efficient filters when applying the algorithm, thereby bringing better and more stable steganalysis security to the present invention.

[0047] (2) The large amplitude priority criterion proposed in the existing DFR method can prevent the modification of small amplitude sampling points to a certain extent, but it will conflict with the complexity priority criterion of content adaptive steganography to a certain extent. The present invention proposes a small amplitude suppression criterion. After applying it to the complexity priority criterion, the small amplitude sampling points are set as "wet points", which is committed to accurately preventing the small amplitude sampling points from being modified, while not affecting the embedded modification of other amplitude sampling points. This makes the encrypted audio obtained by the present invention more difficult to be detected by the steganalyzer, thereby improving the steganographic security of the present invention, and has the advantages of higher steganographic security and better encrypted audio quality.

[0048] (3) The present invention surpasses the most advanced similar algorithms in terms of security and audio quality of encrypted audio, achieves better performance, and has better steganographic imperceptibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1Schematic diagram of the flow of the time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification of the present invention;

[0050] Figure 2 A schematic diagram of the implementation process framework for constructing a steganographic distortion cost function in the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0052] Example 1

[0053] like Figure 1 As shown, this embodiment provides a time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification, including the following steps:

[0054] S1: Acquire carrier audio, perform segmentation preprocessing on the carrier audio, and cut the carrier audio into audio segments of a set length. In this embodiment, the length of each audio segment is preferably 1 second.

[0055] S2: Slice the secret information bits to be embedded equally according to the number of audio segments to obtain secret information bit segments;

[0056] S3: For each audio segment, a stego-distortion cost function is first constructed to calculate its stego-distortion cost. Then, the secret information bit segments are embedded into the segment using the STC stego-coding method based on the minimum distortion steganography framework to obtain multiple encrypted audio segments.

[0057] S4: All the encrypted audio clips are spliced ​​together to obtain the final encrypted audio.

[0058] In this embodiment, the stego-embedding distortion cost corresponding to the carrier is calculated based on the construction of a stego-distortion cost function. The construction process of the stego-distortion cost function is as follows: first, based on the complexity priority criterion, a stego-distortion cost function based on the generalized audio intrinsic energy is proposed. On this basis, a distortion cost design criterion for micro-amplitude suppression modification is proposed to supplement it.

[0059] In this embodiment, the design of the steganalysis distortion cost function based on generalized audio intrinsic energy specifically includes:

[0060] The energy of a digital signal can be defined as the sum of the squares of the amplitudes of all its frequency components. For a one-dimensional audio signal, its frequency components can be easily obtained using one-dimensional frequency transform functions, such as DCT (discrete cosine transform), DFT (discrete Fourier transform), DWT (discrete wavelet transform), etc.

[0061] like Figure 2 As shown, this embodiment first takes a sliding window SW of size n (n>1) n To intercept a local carrier audio signal and express it as x l =[x1,x2...x n ], where n is usually an odd number. For the extraction of the carrier audio frequency component, this embodiment preferably uses DCT transformation to achieve it. Then, by performing the local carrier audio signal x l By performing DCT transformation, n different frequency components can be obtained, namely F1, F2...F n Assume that the i-th frequency component F i The intensity (amplitude) is f i , then referring to the definition of digital signal energy, the local carrier audio signal x l Inner energy It can be expressed as:

[0062]

[0063] Embedding modifications in the high-frequency region (high-complexity region) of the carrier are generally safer than embedding modifications in the low-frequency region (low-complexity region). This means that when the signal energy in the above energy formula is equal, the high-frequency region is theoretically more suitable for embedding. Therefore, when calculating the intrinsic energy of the audio to design the steganographic distortion cost, different weights w should be assigned to different frequency components in formula (1). The principle is that the higher the frequency, the greater the weight. In addition, the steganographic distortion cost is generally related to the content complexity of the carrier, but not to the content of the carrier (i.e., the amplitude of the carrier signal). The results of the DFR_based method also confirm that the audio signal amplitude is not very suitable for the design of its steganographic distortion cost. Therefore, it is necessary to remove the direct current frequency component f1 in the formula to eliminate the influence of the audio amplitude on its intrinsic energy calculation. On the other hand, in order to facilitate the adjustment of the weight w, an exponential adjustment coefficient p is introduced for the weight w; so far, in the solution proposed in this embodiment, the local carrier audio signal x l Intrinsic energy for steganography The construction can be further expressed in the following generalized form:

[0064]

[0065] Without loss of generality, in order to meet the needs of audio steganography distortion cost function design, in the solution proposed in this embodiment, will be considered as x l The center point x cThe complexity index of the carrier audio is designed to be inversely proportional to its complexity. According to this paradigm, the distortion cost function of the proposed time-domain audio steganography based on generalized audio intrinsic energy is defined as:

[0066]

[0067] Among them, ρ c is x c The distortion cost, λ is to prevent ρ c The equilibrium constant is too small, and ε is the stability constant. Slide SW for each sampling point in turn n Window, and then use the above formula to get the distortion cost of all sampling points in each audio carrier sample. When calculating the intrinsic steganographic energy of the audio, exclude the influence of the DC frequency component, that is, discard f1;

[0068] Note that mirror padding will be used to obtain x if necessary. l Since the audio steganography distortion cost function proposed in the present invention is designed based on the generalized audio intrinsic energy, it is named GAIE (Generalized Audio Intrinsic Energy) method for the convenience of description.

[0069] In the above formula, the stability constant ε is taken as 10 -2 , the weight w corresponding to the i-th frequency component i Set to w i =i(2≤i≤n). Considering the exponential weight w i p The equilibrium constant λ in the formula is set to n p *10 6 As for the parameter selection of the sliding window size n, experiments have found that as n increases, the steganographic security performance of the present invention will also increase accordingly. When n reaches 169, the performance tends to be stable. Therefore, the value of n is preferably fixed at 169. In addition, experiments have shown that the optimal value of the parameter p is closely related to the data set. After analysis, the parameter p mainly acts on the weight of the frequency component. Therefore, an attempt is made to determine the parameter p by the percentage of signal energy occupied by the high-frequency components in each carrier (pseh f). The specific definition of psehf is as follows:

[0070]

[0071] Among them, k (k≥2) is the selection threshold of high-frequency components. Experiments show that the optimal value of k is 6. is the frequency F of all local signal segments in each vector i The average strength of the sample, DC frequency also needs to be excluded. The larger the psehf of the sample, the greater the proportion of its high-frequency components. Experiments have shown that when the psehf is larger, a relatively large p can achieve better steganalytic security. Therefore, the specific correspondence between the psehf and the parameter p is constructed as follows:

[0072]

[0073] In this embodiment, the design of the steganalysis distortion cost function based on generalized audio intrinsic energy and micro-amplitude suppression modification specifically includes:

[0074] To address the drawback of the LAF criterion in the DFR algorithm, the present invention proposes a micro-amplitude suppression criterion (MAS). By combining it with the complexity priority criterion, that is, after applying the complexity priority criterion, the micro-amplitude sampling points are set as "wet points" (representing unusable points), aiming to accurately prevent the micro-amplitude sampling points from being modified while not affecting the embedding modification of other amplitude sampling points. By combining this MAS criterion with the above GAIE, the key technology of the present invention can be obtained - the steganalysis distortion cost function based on generalized audio intrinsic energy and micro-amplitude suppression modification (GAIE_MAS), which is as follows:

[0075]

[0076] Among them, |x c | is the sampling point x c The amplitude of the sample point is suppressed, T is the amplitude threshold of the modified sampling point. ψ is the distortion cost of the wet point under the MAS criterion. In this embodiment, the distortion cost of the wet point under the MAS criterion is preferably set to 10 20 That is, when the amplitude of a sampling point is less than the threshold, the distortion cost will be very large to prevent the point from being modified.

[0077] Existing DFR methods use derivative filter residuals to characterize carrier complexity and construct their steganodistortion cost functions. However, the steganographic security performance of this method is often heavily dependent on the selection of efficient filters. Therefore, this invention abandons the idea of ​​using carrier signal residuals to characterize carrier complexity and instead utilizes the carrier's intrinsic energy to characterize carrier complexity. Based on this, a new steganodistortion cost function is designed. This eliminates the need for carefully selecting efficient filters when applying the algorithm, thereby providing the present invention with superior and more stable steganodistortion security.

[0078] In addition, the large amplitude priority criterion proposed in the existing DFR method can prevent the modification of small amplitude sampling points to a certain extent, but it will conflict with the complexity priority criterion of content-adaptive steganography to a certain extent. For this reason, the present invention proposes a micro-amplitude suppression criterion (Micro-Amplitude-Suppression, MAS). After applying it to the complexity priority criterion, the small amplitude sampling points are set as "wet points", which is committed to accurately preventing the small amplitude sampling points from being modified, while not affecting the embedded modification of other amplitude sampling points. This makes the encrypted audio obtained by the present invention more difficult to be detected by the steganalyzer, thereby improving the steganographic security of the present invention. The present invention has the advantages of higher steganographic security and better encrypted audio quality.

[0079] In order to verify that the steganography method of the present invention has better steganography security, tests are conducted on four audio datasets: CMU_ARCTIC (CMU), LJSpeech (LJ), TED-LIUM corpus (TED), and Arabic Speech (ARABIC);

[0080] For each of the four datasets mentioned above, we first randomly select some audio clips from them, then crop them into a series of shorter clips of 1 second in length and store them in WAV format. 10,000 1-second audio clips are selected from each dataset for experiments, thus forming four audio datasets.

[0081] During testing, all steganographic schemes were simulated under their corresponding load-distortion bounds, with relative loads α∈{0.1, 0.2, 0.3, 0.4, 0.5} bps. For the design of the steganalyzer, the currently optimal steganalysis feature TMF was used to evaluate the empirical security performance of the steganographic scheme, and FLD was used to train the ensemble classifier. Half of the original carrier / secret carrier audio pairs in the dataset will be used to train the classifier, and the other half will be used as a test set. The security performance is defined as the average classification error probability under the same prior conditions obtained by performing 20 random tests on the test set, denoted as Average classification error probability The closer it is to 50%, the better the security of the steganography scheme.

[0082] To demonstrate the advantages of the proposed GAIE_MAS scheme, we compared its security performance with existing state-of-the-art steganography schemes, AAC and DFR. Tables 1 and 2 below show the performance comparisons of the tested steganography schemes on four datasets. As can be seen from the tables, the performance of AAC and DFR methods is unstable and fluctuates significantly across different datasets. The comprehensive scheme GAIE_MAS, however, achieved optimal or comparable performance across nearly all datasets.

[0083] Table 1 Security performance of steganography schemes against TMF feature detection on CMU and LJ datasets contrast

[0084]

[0085] Table 2 Security performance of steganography schemes against TMF feature detection on TED and ARABIC datasets contrast

[0086]

[0087] In addition to steganographic security, imperceptibility is also an important metric for evaluating steganographic quality. To compare the imperceptibility of different time-domain audio steganographic schemes, the Perceptual Evaluation of Audio Quality (PEAQ) metric was introduced. This is an objective evaluation standard for audio perception recommended by ITU-R BS.1387. It first uses the subjective perception characteristics of the human ear to calculate the masking threshold and distortion threshold of the audio signal. An artificial neural network is then used to integrate the evaluation parameter ODG (Object Difference Grade). A higher ODG score indicates better audio quality. This example selected 2,000 audio clips from CMU and LJ, respectively, and then compared the audio quality of the tested steganographic schemes under different loads. As shown in Table 3 below, the ODG score comparison results show that the GAIE_MAS scheme proposed in this invention has the highest ODG score, meaning it has the best steganographic imperceptibility.

[0088] Table 3 Comparison of ODG scores of steganography schemes tested on CMU and LJ

[0089]

[0090] Example 2

[0091] This embodiment provides a time-domain audio steganography system based on generalized audio intrinsic energy and micro-amplitude suppression modification, which is used to implement the time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification described in the first embodiment. The system includes: a carrier audio acquisition module, a carrier audio preprocessing module, a secret information bit preprocessing module, a stego-distortion cost function construction module, a stego-distortion cost calculation module, a stego-encoding module, and an audio splicing module.

[0092] In this embodiment, the carrier audio acquisition module is used to acquire the carrier audio;

[0093] In this embodiment, the carrier audio preprocessing module is used to perform segmentation preprocessing on the carrier audio, cutting the carrier audio into audio segments of a set length;

[0094] In this embodiment, the secret information bit preprocessing module is used to equally slice the secret information bits to be embedded according to the number of audio segments to obtain secret information bit segments;

[0095] In this embodiment, the steganalysis distortion cost function construction module is used to construct a steganalysis distortion cost function based on generalized audio intrinsic energy based on the complexity priority criterion. The steganalysis distortion cost function based on generalized audio intrinsic energy is combined with the distortion cost design criterion of micro-amplitude suppression modification to obtain the final steganalysis distortion cost function.

[0096] In this embodiment, the steganalysis distortion cost calculation module is used to calculate the steganalysis distortion cost of each audio clip;

[0097] In this embodiment, the stego-coding module is used to embed a secret information bit segment into each audio segment based on the STC stego-coding method of the minimum distortion steganography framework, thereby obtaining a plurality of secret audio segments;

[0098] In this embodiment, the audio splicing module is used to splice all encrypted audio segments to obtain the final encrypted audio.

[0099] Example 3

[0100] This embodiment provides a storage medium, which can be a storage medium such as ROM, RAM, disk, or CD. The storage medium stores one or more programs. When the program is executed by a processor, the time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification of embodiment 1 is implemented.

[0101] Example 4

[0102] This embodiment provides a computing device, which can be a desktop computer, a laptop computer, a smart phone, a PDA handheld terminal, a tablet computer, or other terminal device with a display function. The computing device includes a processor and a memory, and the memory stores one or more programs. When the processor executes the program stored in the memory, the time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification of Example 1 is implemented.

[0103] The above embodiments are preferred implementations of the present invention, but the implementations of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification, characterized in that: The steps include: Obtain carrier audio, perform segmentation preprocessing on the carrier audio, and cut the carrier audio into audio segments of a set length; Slicing the secret information bits to be embedded into equal parts according to the number of audio segments to obtain secret information bit segments; Based on the complexity priority criterion, a steganalysis distortion cost function based on generalized audio intrinsic energy is constructed. The steganalysis distortion cost function based on generalized audio intrinsic energy is combined with the distortion cost design criterion of micro-amplitude suppression modification to obtain the final steganalysis distortion cost function. Construct a steganographic distortion cost function based on the generalized audio intrinsic energy, which is specifically expressed as: Among them, ρ c represents the distortion cost of the center point of the local carrier audio signal, and λ represents the distortion cost used to prevent ρ c The equilibrium constant is too small, ε represents the stability constant, n represents the number of different carrier audio frequency components, f i represents the amplitude of the i-th carrier audio frequency component, w i represents the weight corresponding to the i-th carrier audio frequency component, p represents the exponential adjustment coefficient, Represents the local carrier audio signal x l Intrinsic energy used for steganography; when calculating the intrinsic steganographic energy of audio, the influence of the DC frequency component is excluded, that is, f1 is discarded; Combining the steganalysis distortion cost function based on generalized audio intrinsic energy with the distortion cost design criterion of slight amplitude suppression modification, the final steganalysis distortion cost function is obtained, which is specifically expressed as: Among them, |x c | is the center point x of the local carrier audio signal c The amplitude of , T is the amplitude threshold of the modified sampling point, ψ is the distortion cost of the wet point under the micro-amplitude suppression criterion; The steganalysis distortion cost of each audio segment is calculated, and the STC steganographic coding method based on the minimum distortion steganography framework is used to embed the secret information bit segment into each audio segment, thereby obtaining multiple encrypted audio segments. All the encrypted audio clips are spliced ​​together to obtain the final encrypted audio.

2. The time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification according to claim 1 is characterized in that: Construct a steganographic distortion cost function based on generalized audio intrinsic energy, specifically including: Obtaining a local carrier audio signal, and obtaining different carrier audio frequency components based on spectrum transformation; Different weights are set for different carrier audio frequency components, and the weights are used to set exponential adjustment coefficients to calculate the intrinsic energy of the local carrier audio signal for steganography; The intrinsic energy of the local carrier audio signal used for steganography is used as the complexity index of the center point of the local carrier audio signal. The steganographic distortion cost of the carrier audio is inversely proportional to the complexity. A steganographic distortion cost function based on the generalized audio intrinsic energy is constructed.

3. The time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification according to claim 2 is characterized in that: The local carrier audio signal is obtained by setting a sliding window to intercept or using mirror filling to obtain the local carrier audio signal.

4. The time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification according to claim 2 is characterized in that: The spectrum transformation adopts any one of discrete cosine transform, discrete Fourier transform or discrete wavelet transform.

5. The time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification according to claim 2 is characterized in that: The index adjustment factor is expressed as: Among them, psehf represents the energy percentage of high-frequency components in each carrier, k represents the selection threshold of high-frequency components, k≥2, is the average amplitude of all carrier audio frequency components in each carrier.

6. A time-domain audio steganography system based on generalized audio intrinsic energy and micro-amplitude suppression modification, characterized in that: include: Carrier audio acquisition module, carrier audio preprocessing module, secret information bit preprocessing module, steganalysis distortion cost function construction module, steganalysis distortion cost calculation module, steganalysis encoding module, audio splicing module; The carrier audio acquisition module is used to acquire carrier audio; The carrier audio preprocessing module is used to perform segmentation preprocessing on the carrier audio, cutting the carrier audio into audio segments of a set length; The secret information bit preprocessing module is used to equally slice the secret information bits to be embedded according to the number of audio segments to obtain secret information bit segments; The steganalysis distortion cost function construction module is used to construct a steganalysis distortion cost function based on generalized audio intrinsic energy based on a complexity priority criterion, and combine the steganalysis distortion cost function based on generalized audio intrinsic energy with the distortion cost design criterion of micro-amplitude suppression modification to obtain a final steganalysis distortion cost function; Construct a steganographic distortion cost function based on the generalized audio intrinsic energy, which is specifically expressed as: Among them, ρ c represents the distortion cost of the center point of the local carrier audio signal, and λ represents the distortion cost used to prevent ρ c The equilibrium constant is too small, ε represents the stability constant, n represents the number of different carrier audio frequency components, f i represents the amplitude of the i-th carrier audio frequency component, w i represents the weight corresponding to the i-th carrier audio frequency component, p represents the exponential adjustment coefficient, Represents the local carrier audio signal x l Intrinsic energy used for steganography; when calculating the intrinsic steganographic energy of audio, the influence of the DC frequency component is excluded, that is, f1 is discarded; Combining the steganalysis distortion cost function based on generalized audio intrinsic energy with the distortion cost design criterion of slight amplitude suppression modification, the final steganalysis distortion cost function is obtained, which is specifically expressed as: Among them, |x c | is the center point x of the local carrier audio signal c The amplitude of the sample point is suppressed, T is the amplitude threshold of the modified sampling point, ψ is the distortion cost of the wet point under the micro-amplitude suppression criterion, w i represents the weight corresponding to the i-th carrier audio frequency component, p represents the exponential adjustment coefficient, n represents the number of different carrier audio frequency components, ε represents the stability constant, ρ c Represents the distortion cost of the center point of the local carrier audio signal; The steganalysis distortion cost calculation module is used to calculate the steganalysis distortion cost of each audio clip; The stego-coding module is used to embed a secret information bit segment into each audio segment based on the STC stego-coding method of the minimum distortion steganographic framework to obtain multiple secret audio segments; The audio splicing module is used to splice all the encrypted audio segments to obtain the final encrypted audio.

7. A computer-readable storage medium storing a program, characterized in that: When the program is executed by a processor, the time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification as described in any one of claims 1 to 5 is implemented.

8. A computer device comprising a processor and a memory for storing a program executable by the processor, characterized in that When the processor executes the program stored in the memory, it implements the time-domain audio steganography method based on generalized audio intrinsic energy and micro-amplitude suppression modification as described in any one of claims 1-5.

Images