Construction method of deep learning-based streaming encryption algorithm

Abstract

The application discloses a kind of construction methods of streaming encryption algorithm based on deep learning;Increase encrypted circle, the edge of the division encrypted circle is 360 arc, the relationship between the arc of encrypted circle and character is matched, and its relationship is stored in matching network;In order to increase the accuracy of matching network, detection network is increased, the parameter of matching network is adjusted, and the output value of matching network is also adjusted, so that more accurate matching data is found;So as to only through matching network, detection network, the size of the center of circle and corresponding encrypted circle, accurate encrypted second prediction position vector can be obtained, and then through decryption network, and when training again, decryption network, detection network and matching network are trained together.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to a method for constructing a streaming encryption algorithm based on deep learning. Background Technology

[0002] Currently, stream encryption (CK) is a permutation-based, multi-key, symmetric encryption method. Most symmetric encryption algorithms are variations of block encryption. This means that encryption requires aligning preceding data, and data within a block depends on each other. During block generation, previously generated ciphertext is used in the generation of the next ciphertext, requiring a complete block for decryption. This necessitates data alignment for random read / write operations. Stream encryption (CK), however, only performs permutation encryption on characters. Then, using a multi-key algorithm, it generates a pseudo-random stream. The size of the pseudo-random stream is the same as the plaintext. The algorithm can directly calculate the value of the pseudo-random stream at a specific position, thus most effectively supporting random read / write operations and allowing data to be stored directly in ciphertext form on the storage medium.

[0003] Because the number and range of characters in stream encryption are fixed, the encryption method is not as complex as block encryption. Therefore, it is necessary to increase the confidentiality of the encryption to make it difficult for anyone other than the decryptor to decrypt it. Summary of the Invention

[0004] The purpose of this invention is to provide a method for constructing a deep learning-based streaming encryption algorithm to solve the aforementioned problems in the prior art, including:

[0005] Obtain the data to be encrypted and the set of plaintext characters; the data to be encrypted represents data that needs to be encrypted composed of multiple characters; the set of plaintext characters includes multiple characters that can constitute the data to be encrypted;

[0006] Based on the plaintext character set, an encrypted circle is constructed, and a matching network is trained; the matching network represents a network that can match the values ​​in the plaintext character set with the arcs of the encrypted circle one-to-one.

[0007] The characters in the data to be encrypted are input into a trained matching network to obtain a set of predicted positions; multiple characters in the data to be encrypted correspond to multiple sets of predicted positions; the set of predicted positions represents the position of the data to be encrypted on the arc corresponding to the encrypted circle.

[0008] By detecting the network, multiple predicted position sets are adjusted based on the radius of the encryption circle to obtain a second predicted position vector; the first value of the second predicted position vector represents the radius of the encryption circle, and the other values ​​represent the data to be encrypted.

[0009] Optional, also includes:

[0010] The first value of the second predicted position vector is used as the decryption radius, and the values ​​of the second predicted position vector other than the first value are used as the vector to be decrypted.

[0011] The length of the arc corresponding to the decryption radius is used as the segmentation length;

[0012] The vector to be decrypted is divided according to the segment length to obtain multiple sets to be decrypted;

[0013] The set to be decrypted is input into the decryption network to obtain decrypted characters; multiple sets to be decrypted correspond to multiple decrypted characters.

[0014] The decrypted characters are arranged in the order of the encrypted circle division to obtain the decrypted data; the decrypted data is equal to the data to be decrypted.

[0015] Optionally, the step of constructing an encrypted circle and training a matching network based on the plaintext character set includes:

[0016] Obtain an encrypted circle image; an encrypted circle is drawn in the encrypted circle image;

[0017] Use one position on the edge of the encrypted circle as the starting position;

[0018] Divide the edge of the encrypted circle into 360 equal parts from the starting position to obtain 360 arcs;

[0019] Following a clockwise direction, obtain the position of the arc in the encrypted circle image to get the arc position set; obtain 360 arc position sets corresponding to 360 arcs;

[0020] The values ​​in the plaintext character set are matched with the arc position set to obtain a matching relationship set; one value in the plaintext character set corresponds to one arc position set.

[0021] A matching network is trained based on the set of matching relationships.

[0022] Optionally, the step of adjusting multiple predicted position sets based on the radius of the encrypted circle through the detection network to obtain a second predicted position vector includes:

[0023] The values ​​in the predicted position set are marked to obtain the predicted arc position image; the size of the predicted arc position image is equal to the size of the encrypted circle image; multiple predicted position sets correspond to multiple predicted arc position images;

[0024] The predicted arc position image is input into the detection network to determine the state of the encrypted arc, and the predicted angle and predicted radius are obtained; the predicted angle represents the central angle of the arc corresponding to the predicted position set; the predicted radius represents the radius of the circle corresponding to the arc corresponding to the predicted position set.

[0025] Multiple sets of predicted locations correspond to multiple predicted angles and multiple predicted radii;

[0026] Clustering multiple prediction angles yields cluster angles; clustering multiple prediction radii yields cluster radii.

[0027] Based on the clustering radius, the circle is reconstructed to obtain a reconstructed circle image;

[0028] Based on clustering, the predicted arc position image is matched with the reconstructed circle image to obtain a second predicted position set; multiple predicted arc position images correspond to multiple second predicted position sets.

[0029] Optionally, training the matching network based on the set of matching relationships includes:

[0030] The values ​​in the plaintext character set are input into the matching network to detect the position of the corresponding arc of the encryption circle, thus obtaining the predicted encryption circle position set.

[0031] The set of predicted encryption circle positions contains multiple predicted encryption circle positions; the index of the predicted encryption circle position indicates the clockwise order in which the predicted encryption circle positions are obtained.

[0032] The matching network is used to store the matching relationship between values ​​in the plaintext character set and the arc of the encrypted circle;

[0033] The loss is calculated by comparing the predicted encrypted circle position set with the arc position set in the matching relationship set, and the matching network is trained.

[0034] Optionally, the step of matching the predicted arc position image and the reconstructed circle image based on clustering angles to obtain a second predicted position set includes:

[0035] The predicted arc position image and the reconstructed circle image are superimposed to form an overlapping image;

[0036] The predicted arc position image and the center of the reconstructed circle are combined to form a predicted graphic.

[0037] Based on the center of the reconstructed circle, the clustering angle, and the clustering radius, construct the clustering sector;

[0038] Using the center of the reconstructed circle as the rotation center, rotate the clustered sectors sequentially;

[0039] The area difference between the rotated cluster sector and the predicted pattern is calculated to obtain the difference area;

[0040] Multiple rotations correspond to obtaining multiple phase difference areas;

[0041] The smallest phase difference area among multiple phase difference areas is used as the predicted area;

[0042] Add the position of the arc of the cluster sector corresponding to the predicted area to the second predicted position set.

[0043] Optionally, the training method for the detection network includes:

[0044] Obtain the annotation angle and annotation radius; the annotation angle is 1; the annotation radius is the radius of the encrypted circle;

[0045] The loss is calculated at multiple positions corresponding to the reconstructed circle image and the encrypted circle image to obtain the first loss value;

[0046] The loss is calculated by comparing the predicted angle and the labeled angle to obtain the second loss value;

[0047] The loss is calculated by comparing the predicted radius with the labeled radius to obtain the third loss value.

[0048] Optionally, the input to the matching network is a character in the data to be encrypted;

[0049] The output of the matching network is the input of the detection network;

[0050] The output of the detection network is the input of the decryption network;

[0051] The output of the decryption network is the decryption character corresponding to the character in the data to be encrypted.

[0052] Optionally, the loss between the decrypted character and the corresponding character in the data to be encrypted can be calculated, and the decryption network, detection network, and matching network can be trained together.

[0053] Optionally, the length of the second predicted position vector increases as the radius of the encryption circle increases.

[0054] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0055] This invention adds an encryption circle to increase encryption complexity and make decryption difficult without a fixed-parameter decryption network. The encryption circle is divided into 360 arcs, and the relationship between these arcs and characters is matched and stored in a matching network. To improve the accuracy of the matching network, a detection network is added. The parameters and output values ​​of the matching network are adjusted to find more accurate matching data. Thus, using only the matching network, detection network, circle center, and the corresponding encryption circle size, an accurate second predicted position vector can be obtained. This vector is then used in a decryption network, and during retraining, the decryption, detection, and matching networks are trained together, enabling the trained decryption network to perform decryption more accurately. Attached Figure Description

[0056] Figure 1 This is a flowchart of a method for constructing a deep learning-based streaming encryption algorithm provided in an embodiment of the present invention.

[0057] Figure 2 This is a schematic diagram of the encryption circle in the flowchart of a method for constructing a deep learning-based streaming encryption algorithm provided in an embodiment of the present invention.

[0058] Figure 3 This is a schematic diagram of the overlapping images in the flowchart of a method for constructing a deep learning-based streaming encryption algorithm provided in an embodiment of the present invention. Detailed Implementation

[0059] The present invention will now be described in detail with reference to the accompanying drawings.

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment of the invention provides a method for constructing a deep learning-based streaming encryption algorithm, the method comprising:

[0062] S101: Obtain the data to be encrypted and the set of plaintext characters; the data to be encrypted represents multiple characters that constitute the data to be encrypted; the set of plaintext characters includes multiple characters that can constitute the data to be encrypted.

[0063] The plaintext character set includes 256 characters in the range [0x00, 0xff].

[0064] S102: Based on the plaintext character set, construct an encrypted circle and train a matching network; the matching network represents a network that can correspond one-to-one with the values ​​in the plaintext character set and the arcs of the encrypted circle.

[0065] S103: Input the characters in the data to be encrypted into the trained matching network to obtain a set of predicted positions; multiple characters in the data to be encrypted correspond to multiple sets of predicted positions; the set of predicted positions represents the position of the data to be encrypted on the corresponding arc on the encryption circle.

[0066] S104: By detecting the network, based on the radius of the encryption circle, adjust multiple predicted position sets to obtain a second predicted position vector; the first value of the second predicted position vector represents the radius of the encryption circle, and the other values ​​represent the data to be encrypted.

[0067] In this embodiment, the radius of the encryption circle is fixed. A larger radius indicates a larger arc length of the arc used for matching, and thus a longer length of the second predicted position vector.

[0068] By using the above method to represent characters with arcs, even if there are minor areas where prediction is inaccurate, using the values ​​corresponding to multiple arcs to represent a single character allows for holistic judgment based on shape and position. This ensures accurate prediction even when some parts are inaccurate, thus addressing the inaccuracy issues inherent in the network.

[0069] Optional, also includes:

[0070] The first value of the second predicted position vector is used as the decryption radius, and the values ​​of the second predicted position vector other than the first value are used as the vector to be decrypted.

[0071] The decryption radius is retained during the transmission process.

[0072] The length of the arc corresponding to the decryption radius is used as the segmentation length;

[0073] The vector to be decrypted is divided according to the segment length to obtain multiple sets to be decrypted;

[0074] The set to be decrypted is input into the decryption network to obtain decrypted characters; multiple sets to be decrypted correspond to multiple decrypted characters.

[0075] In this embodiment, the decryption network is a convolutional neural network (CNN).

[0076] The decrypted characters are arranged in the order of the encrypted circle division to obtain the decrypted data; the decrypted data is equal to the data to be decrypted.

[0077] Optionally, the step of constructing an encrypted circle and training a matching network based on the plaintext character set includes:

[0078] Obtain an encrypted circle image; an encrypted circle is drawn in the encrypted circle image.

[0079] The encrypted circle image is a binary image; the center of the encrypted circle is at the midpoint of the encrypted circle image.

[0080] Use one position on the edge of the encrypted circle as the starting position;

[0081] Divide the edge of the encrypted circle into 360 equal parts from the starting position to obtain 360 arcs.

[0082] The arc refers to a segment of an arc on the edge of the encrypted circle.

[0083] Since it's character encryption, the range that a single character can represent is exactly [0x00, 0xff]. We represent the plaintext character to be encrypted with C, meaning C must be a number between [0x00, 0xff]. Therefore, dividing the encryption circle into 360 equal parts, 360 > 256, satisfies the condition that the character can be represented by the edge of the encryption circle.

[0084] The schematic diagram of the encryption circle is shown below. Figure 2 As shown.

[0085] The values ​​in the plaintext character set are matched with the arc to obtain a set of matching relationships.

[0086] In this context, one value in the plaintext character set corresponds to one arc, and one arc necessarily corresponds to one value in the plaintext character set.

[0087] In this embodiment, one value from the plaintext character set is sequentially matched with an arc in a clockwise direction. For example, the character 0x00 corresponds to the arc that is rotated 1 degree clockwise from the starting position, and 0x01 corresponds to the arc that is rotated 2 degrees clockwise from the starting position.

[0088] A matching network is trained based on the set of matching relationships.

[0089] Optionally, the step of adjusting multiple predicted position sets based on the radius of the encrypted circle through the detection network to obtain a second predicted position vector includes:

[0090] The values ​​in the predicted position set are marked to obtain the predicted arc position image; the size of the predicted arc position image is equal to the size of the encrypted circle image; multiple predicted position sets correspond to multiple predicted arc position images;

[0091] The image of the predicted arc position is input into the detection network to determine the state of the encrypted arc, and the predicted angle and predicted radius are obtained; the predicted angle represents the central angle of the arc corresponding to the predicted position set; the predicted radius represents the radius of the circle corresponding to the arc corresponding to the predicted position set.

[0092] In this embodiment, the detection network is a fully connected neural network (FCN).

[0093] Multiple sets of predicted locations correspond to multiple predicted angles and multiple predicted radii;

[0094] Clustering multiple prediction angles yields cluster angles; clustering multiple prediction radii yields cluster radii.

[0095] The clustering angle and clustering radius are cluster centers. Clustering the predicted angle and predicted radius is to obtain the same angle and radius as the encrypted circle.

[0096] In this embodiment, the k-means method is used for clustering.

[0097] Based on the cluster radius, the circle is reconstructed to obtain a reconstructed circle image.

[0098] In order to determine the accuracy of the matching network and the detection network, the detection network is used to train the matching network to obtain a more accurate network for matching arcs and characters.

[0099] Specifically, a circle is drawn with the center of the encrypted circle as its center and the cluster radius as its radius, resulting in a reconstructed circle image. The size of the reconstructed circle image is equal to the size of the encrypted circle image.

[0100] Based on clustering, the predicted arc position image is matched with the reconstructed circle image to obtain a second predicted position set; multiple predicted arc position images correspond to multiple second predicted position sets.

[0101] In this case, because there are slight differences between the values ​​in the reconstructed circle and the predicted arc position set, the predicted arc position set is replaced by the second predicted position set on the reconstructed circle.

[0102] Optionally, the training method for the matching network includes:

[0103] The values ​​in the plaintext character set are input into the matching network, and the positions of the corresponding encrypted circle arcs are detected to obtain the predicted encrypted circle position set.

[0104] In this embodiment, the values ​​in the plaintext character set are matched with corresponding arcs. Since one value in the plaintext character set corresponds to one arc, the arc can be used to identify the corresponding character.

[0105] The number of output neurons in the matching network is equal to the number of elements in the set of arc positions. Because the edges of the encrypted circle are bisected, the arc lengths are the same, so the number of elements in the 360 ​​arc position sets is the same.

[0106] In this embodiment, the matching network is a convolutional neural network (CNN).

[0107] The set of predicted encryption circle positions contains multiple predicted encryption circle positions; the index of the predicted encryption circle position indicates the clockwise order in which the predicted encryption circle positions are obtained.

[0108] The matching network is used to store the matching relationship between values ​​in the plaintext character set and the arc of the encrypted circle;

[0109] The loss is calculated by applying the predicted encrypted circle position set and the arc position set to train the matching network.

[0110] In this embodiment, the values ​​corresponding to the indices of the predicted encrypted circle position set and the arc position set are averaged after calculating the loss to obtain the loss value.

[0111] In this embodiment, the cross-entropy loss function is used to calculate the loss value. The backward-trained matching network is then used.

[0112] Optionally, the step of matching the predicted arc position image and the reconstructed circle image based on clustering angles to obtain a second predicted position set includes:

[0113] The predicted arc position image and the reconstructed circle image are superimposed to form an overlapping image.

[0114] In this embodiment, the overlapping image is as follows: Figure 3 As shown.

[0115] The predicted arc position image and the center of the reconstructed circle are combined to form a predicted graphic.

[0116] Based on the center of the reconstructed circle, the clustering angle, and the clustering radius, construct the clustering sector;

[0117] Using the center of the reconstructed circle as the rotation center, rotate the clustered sectors sequentially;

[0118] The area difference between the rotated cluster sector and the predicted pattern is calculated to obtain the difference area;

[0119] Multiple rotations correspond to obtaining multiple phase difference areas;

[0120] The smallest phase difference area among multiple phase difference areas is used as the predicted area;

[0121] Add the positions of the arcs of the clustered sectors corresponding to the predicted areas to the second predicted position set.

[0122] In this embodiment, Figure 3The character corresponding to the predicted arc position image of the set of predicted arc positions represented in the image is "a".

[0123] Optionally, the training method for the detection network includes:

[0124] Obtain the annotation angle and annotation radius; the annotation angle is 1; the annotation radius is the radius of the encrypted circle;

[0125] The loss is calculated at multiple positions corresponding to the reconstructed circle image and the encrypted circle image to obtain the first loss value;

[0126] The loss is calculated by comparing the predicted angle and the labeled angle to obtain the second loss value;

[0127] The loss is calculated by combining the predicted radius and the labeled radius to obtain the third loss value;

[0128] In this embodiment, the cross-entropy loss function is used to obtain the loss value.

[0129] Optionally, the input to the matching network is a character in the data to be encrypted;

[0130] The output of the matching network is the input of the detection network;

[0131] The output of the detection network is the input of the decryption network;

[0132] The output of the decryption network is the decryption character corresponding to the character in the data to be encrypted.

[0133] Optionally, the loss between the decrypted character and the corresponding character in the data to be encrypted can be calculated, and the decryption network, detection network, and matching network can be trained together.

[0134] Optionally, the length of the second predicted position vector increases as the radius of the encryption circle increases.

[0135] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0136] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods and structures have not been shown in detail so as not to obscure the understanding of this specification.

Claims

1. A method for constructing a streaming encryption algorithm based on deep learning, characterized in that, include: Obtain the data to be encrypted and the set of plaintext characters; the data to be encrypted represents data that needs to be encrypted composed of multiple characters; the set of plaintext characters includes multiple characters that can constitute the data to be encrypted; Based on the plaintext character set, an encrypted circle is constructed, and a matching network is trained; the matching network represents a network that can match the values ​​in the plaintext character set with the arcs of the encrypted circle one-to-one. The step of constructing an encrypted circle and training a matching network based on the plaintext character set includes: Obtain an encrypted circle image; an encrypted circle is drawn in the encrypted circle image; Use one position on the edge of the encrypted circle as the starting position; Divide the edge of the encrypted circle into 360 equal parts from the starting position to obtain 360 arcs; Following a clockwise direction, obtain the position of the arc in the encrypted circle image to get the arc position set; obtain 360 arc position sets corresponding to 360 arcs; The values ​​in the plaintext character set are matched with the arc position set to obtain a matching relationship set; one value in the plaintext character set corresponds to one arc position set. Train a matching network based on the set of matching relationships; The characters in the data to be encrypted are input into a trained matching network to obtain a set of predicted positions; multiple characters in the data to be encrypted correspond to multiple sets of predicted positions; the set of predicted positions represents the position of the data to be encrypted on the arc corresponding to the encrypted circle. By detecting the network and adjusting multiple predicted position sets based on the radius of the encryption circle, a second predicted position vector is obtained; the first value of the second predicted position vector represents the radius of the encryption circle, and the other values ​​represent the data to be encrypted. The process of adjusting multiple predicted position sets based on the radius of the encrypted circle through a detection network to obtain a second predicted position vector includes: The values ​​in the predicted position set are marked to obtain the predicted arc position image; the size of the predicted arc position image is equal to the size of the encrypted circle image; multiple predicted position sets correspond to multiple predicted arc position images; The predicted arc position image is input into the detection network to determine the state of the encrypted arc and obtain the predicted angle and predicted radius; the predicted angle represents the central angle of the arc corresponding to the predicted position set; the predicted radius represents the radius of the circle corresponding to the arc corresponding to the predicted position set. Multiple sets of predicted locations correspond to multiple predicted angles and multiple predicted radii; Clustering multiple prediction angles yields cluster angles; clustering multiple prediction radii yields cluster radii. Based on the clustering radius, the circle is reconstructed to obtain a reconstructed circle image; Based on clustering, the predicted arc position image is matched with the reconstructed circle image to obtain a second predicted position set; multiple predicted arc position images correspond to multiple second predicted position sets.

2. The method for constructing a deep learning-based streaming encryption algorithm according to claim 1, characterized in that, Also includes: The first value of the second predicted position vector is used as the decryption radius, and the values ​​of the second predicted position vector other than the first value are used as the vector to be decrypted. The length of the arc corresponding to the decryption radius is used as the segmentation length; The vector to be decrypted is divided according to the segment length to obtain multiple sets to be decrypted; The set to be decrypted is input into the decryption network to obtain decrypted characters; multiple sets to be decrypted correspond to multiple decrypted characters. The decrypted characters are arranged in the order of the encrypted circle division to obtain the decrypted data; the decrypted data is equal to the data to be decrypted.

3. The method for constructing a deep learning-based streaming encryption algorithm according to claim 1, characterized in that, The training of the matching network based on the set of matching relationships includes: The values ​​in the plaintext character set are input into the matching network to detect the position of the corresponding arc of the encryption circle, thus obtaining the predicted encryption circle position set. The set of predicted encryption circle positions contains multiple predicted encryption circle positions; the index of the predicted encryption circle position indicates the clockwise order in which the predicted encryption circle positions are obtained. The matching network is used to store the matching relationship between values ​​in the plaintext character set and the arc of the encrypted circle; The loss is calculated by comparing the predicted encrypted circle position set with the arc position set in the matching relationship set, and the matching network is trained.

4. The method for constructing a deep learning-based streaming encryption algorithm according to claim 1, characterized in that, The second set of predicted positions is obtained by matching the predicted arc position image with the reconstructed circle image based on clustering angle, including: The predicted arc position image and the reconstructed circle image are superimposed to form an overlapping image; The predicted arc position image and the center of the reconstructed circle are combined to form a predicted graphic. Based on the center of the reconstructed circle, the clustering angle, and the clustering radius, construct the clustering sector; Using the center of the reconstructed circle as the rotation center, rotate the clustered sectors sequentially; The area difference between the rotated cluster sector and the predicted pattern is calculated to obtain the difference area; Multiple rotations correspond to obtaining multiple phase difference areas; The smallest phase difference area among multiple phase difference areas is used as the predicted area; Add the position of the arc of the cluster sector corresponding to the predicted area to the second predicted position set.

5. The method for constructing a deep learning-based streaming encryption algorithm according to claim 1, characterized in that, The training method for the detection network includes: Obtain the annotation angle and annotation radius; the annotation angle is 1; the annotation radius is the radius of the encrypted circle; The loss is calculated at multiple positions corresponding to the reconstructed circle image and the encrypted circle image to obtain the first loss value; The loss is calculated by comparing the predicted angle and the labeled angle to obtain the second loss value; The loss is calculated by comparing the predicted radius with the labeled radius to obtain the third loss value.

6. The method for constructing a deep learning-based streaming encryption algorithm according to claim 2, characterized in that, The input to the matching network is the characters in the data to be encrypted; The output of the matching network is the input of the detection network; The output of the detection network is the input of the decryption network; The output of the decryption network is the decryption character corresponding to the character in the data to be encrypted.

7. The method for constructing a deep learning-based streaming encryption algorithm according to claim 2, characterized in that, The loss is calculated by comparing the decrypted character with the corresponding character in the data to be encrypted, and then used to train the decryption network, detection network, and matching network.

8. The method for constructing a deep learning-based streaming encryption algorithm according to claim 1, characterized in that, The length of the second predicted position vector increases as the radius of the encryption circle increases.

Images