CN115378574B - Lightweight dynamic image data encryption method and system - Google Patents
Lightweight dynamic image data encryption method and system Download PDFInfo
- Publication number
- CN115378574B CN115378574B CN202210948717.3A CN202210948717A CN115378574B CN 115378574 B CN115378574 B CN 115378574B CN 202210948717 A CN202210948717 A CN 202210948717A CN 115378574 B CN115378574 B CN 115378574B
- Authority
- CN
- China
- Prior art keywords
- image
- chaotic
- encryption
- sensitive
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000000739 chaotic effect Effects 0.000 claims abstract description 112
- 238000012216 screening Methods 0.000 claims abstract description 45
- 230000035945 sensitivity Effects 0.000 claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 19
- 238000013507 mapping Methods 0.000 claims abstract description 14
- 230000007246 mechanism Effects 0.000 claims abstract description 5
- 239000011159 matrix material Substances 0.000 claims description 24
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 238000001914 filtration Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 13
- 239000013598 vector Substances 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 11
- 230000009467 reduction Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 10
- 230000005540 biological transmission Effects 0.000 claims description 8
- 238000010586 diagram Methods 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 5
- 238000009792 diffusion process Methods 0.000 claims description 5
- 238000013145 classification model Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000011524 similarity measure Methods 0.000 claims description 4
- 238000004138 cluster model Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000002146 bilateral effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/001—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols using chaotic signals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/98—Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
- G06V10/993—Evaluation of the quality of the acquired pattern
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/002—Countermeasures against attacks on cryptographic mechanisms
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/32—Circuits or arrangements for control or supervision between transmitter and receiver or between image input and image output device, e.g. between a still-image camera and its memory or between a still-image camera and a printer device
- H04N1/32101—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title
- H04N1/32144—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title embedded in the image data, i.e. enclosed or integrated in the image, e.g. watermark, super-imposed logo or stamp
- H04N1/32149—Methods relating to embedding, encoding, decoding, detection or retrieval operations
- H04N1/32267—Methods relating to embedding, encoding, decoding, detection or retrieval operations combined with processing of the image
- H04N1/32272—Encryption or ciphering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Computer Security & Cryptography (AREA)
- Computer Networks & Wireless Communication (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a lightweight dynamic image data encryption method and a lightweight dynamic image data encryption system, which are characterized in that structural feature extraction is carried out on images based on an automatic identification image screening mechanism of an image quality screening and feature extraction method, and similarity matching is carried out on the extracted features to realize image screening of specific contents; through image target detection and feature comparison, the screened image is subjected to sensitive area identification, the sensitivity level is calculated according to sensitive objects contained in the image, and relevant sensitive areas meeting the conditions are segmented and extracted; the specific sensitive area is scrambled by using a pixel space scrambling method based on the chaotic sequence, and then a new chaotic sequence is constructed by utilizing dynamic chaotic mapping to diffuse the scrambled image to realize encryption. The lightweight dynamic image data encryption method and system provided by the invention realize lightweight encryption and have better safety performance, and can resist covering attack, differential attack and statistical attack of image data.
Description
Technical Field
The invention belongs to the technical field of computer image encryption, and particularly relates to a lightweight dynamic image data encryption method and system.
Background
With the progress and development of internet technology, image data plays an increasingly important role in the related fields, however, the image data often contains some sensitive information content, and privacy information may be revealed during transmission and storage, especially in some special application scenarios such as medical treatment and finance. Currently, protection of image information generally uses an image encryption process to hide sensitive information or custom private information in an original image to prevent the original image from being stolen or information revealed. However, it is generally difficult for an encryption algorithm designed for text data to securely and efficiently encrypt image data of a large data volume two-dimensional structure and high redundancy.
Furthermore, the ciphertext used to encrypt the image is only related to the key and not to the plaintext, so they are more vulnerable to selective plaintext attacks or known plaintext attacks. Because the chaotic system has stronger sensitivity to initial values and parameters, can generate excellent long-period random sequences, has long-term evolution unpredictability, accords with confusion and diffusion rules of password design, and is therefore often used for image encryption algorithm design and use. However, the multi-dimensional chaotic system has the defects of more parameters, complex algorithm structure, difficult realization, high time consumption and the like, and severely limits the application of the chaotic system in rapid image encryption in real-time image processing, so that the development of an enhanced chaotic system and a lightweight image encryption method with low computational complexity is very necessary.
Disclosure of Invention
The invention aims to provide a lightweight dynamic image data encryption method and system, which are used for solving the problems that a multi-dimensional chaotic system has more parameters, complex algorithm structure, difficult realization, high time consumption and the like, and severely limiting the application of the chaotic system in quick image encryption in real-time image processing.
In order to achieve the above purpose, the present invention provides the following technical solutions: a lightweight dynamic image data encryption method and system includes:
step one: automatic screening of images: extracting structural features of the images based on an automatic image screening mechanism, and matching the similarity of the extracted features to realize the image screening of specific contents;
step two: sensitive object identification: image sensitive object recognition is carried out on the screened image through image target model detection and feature comparison, sensitivity is calculated according to the sensitive object contained in the image, and relevant sensitive areas meeting the conditions are extracted;
step three: dynamic chaotic encryption: scrambling a specific sensitive region by using a regional pixel space scrambling method based on a chaotic sequence, then constructing a new chaotic sequence by using dynamic chaotic encryption mapping, diffusing the scrambled image to realize dynamic chaotic encryption of the image, splicing an encrypted part of the sensitive region and an unencrypted part of a background region, and obtaining an encrypted image for storage or transmission.
Preferably, the method for automatically screening the images specifically comprises the following steps:
(1.1) image quality screening: filtering and denoising the original image, calculating pixel point edge gradients, accumulating and summing the pixel points, and normalizing the edge gradients to screen out clear images;
(1.2) similarity matching: carrying out structural feature extraction on the images meeting the quality requirements through a clustering and classifying algorithm, and screening out images matched with specific targets from the structural feature vectors based on a multi-metric similarity matching algorithm;
the sensitive object identification specifically comprises the following steps:
(2.1) sensitive target detection: identifying, detecting and classifying specific target objects contained in the image;
(2.2) sensitivity calculation: correspondingly calculating a sensitive target object according to a preset sensitivity value;
(2.3) sensitive region extraction: carrying out regional image segmentation extraction on a region which meets the sensitivity requirement of the target object;
the dynamic chaotic encryption comprises the following steps:
(3.1) area pixel scrambling: according to the initial chaotic system, the bytes of the initial image are subjected to preset mess processing, and the obtained array is subjected to dimension reduction by copying the RGB value of each adjacent pixel to obtain a preset mess bitmap.
(3.2) dynamic chaotic encryption: carrying out iterative diffusion on the bitmap pixels subjected to disorder processing according to a sequence of chaotic mapping by dynamically constructing a new chaotic system to obtain a ciphertext matrix;
(3.3) image data stitching: and splicing and integrating the mixed encrypted regional image ciphertext matrix and the unprocessed region of the original image to form the complete data of the target encrypted image.
Preferably, the process of the image quality screening specifically includes:
(1.11) firstly, performing image filtering noise reduction treatment, converting an original image into the filtering noise reduction treatment, filtering noise and smoothing the image, and removing irrelevant details in the image;
(1.12) then carrying out image edge gradient calculation, respectively calculating edge gradients in each pixel point neighborhood, carrying out weighted summation on the edge gradients in each pixel point neighborhood, calculating edge gradient sum of each pixel point, accumulating the edge gradient sum of each pixel point, dividing the edge gradient sum by the number of the pixel points to obtain normalized edge gradients, and screening out a clear image according to the edge gradient value obtained after normalization.
Preferably, the similarity matches include feature clusters and feature classifications,
the feature cluster specifically comprises the following contents:
(1.201) extracting image features of an image to obtain sample image features, and extracting structural features of the image features by using a feature cluster model to obtain at least one image structural feature;
(1.202) clustering the image structural features based on the structural features to obtain at least one clustering result of the structural features;
(1.203) optimizing a preset feature clustering model according to a clustering result until the preset model converges to obtain a feature clustering model and the feature clustering model is used for screening structural features of a target image;
the feature classification specifically comprises:
(1.204) randomly and repeatedly selecting a plurality of image samples from the images, marking feature class labels as training samples, selecting an existing convolutional neural network model for parameter adjustment and optimization, and keeping the model structure and training data different between the convolutional neural network model and the network model for extracting the features;
(1.205) identifying the target image features by using a feature classification model to obtain a structural feature model and the category confidence coefficient thereof, and screening out the features as irrelevant samples if the confidence coefficient of a certain sample on the category to which the sample belongs is very low;
the structural feature extraction comprises the following steps:
(1.21) first, defining structural information as attributes reflecting the structure of objects in a scene independently of brightness and contrast from the viewpoint of image composition, using a mean value as an estimate of brightness, a standard deviation as an estimate of contrast, and covariance as a measure of the degree of structural similarity;
and (1.22) secondly, dividing the image into blocks by utilizing a sliding window, calculating the structural similarity of the corresponding blocks based on pixels in the window in each step, and finally taking the average value as the structural similarity measure of the two images.
Preferably, the specific content of the sensitive target detection is:
according to a preset detection sensitive object list, automatically classifying and identifying target objects contained in the image content to obtain corresponding category information and position information;
extracting features according to the training images of the sensitive objects by using the target detection recognition, and using the obtained feature vectors for training a feature library for target image recognition to obtain a target detection model;
the sensitivity calculation method comprises the following steps:
calculating the sensitivity of the object in the current image containing the object according to the sensitive target detection identification information, wherein the sensitivity depends on the application scene of the image, the target preset value and the semantic relation among the objects in the image scene; the target preset value may employ an initial sensitivity of a predefined given target object in accordance with a general understanding of the image;
the sensitive area extraction process comprises the following steps:
and dividing the original image into a plurality of mutually disjoint areas according to the characteristics of gray scale, color, space texture, geometric shape and the like, including instance segmentation and scene segmentation, so as to obtain an image of the region of the target sensitive object.
Preferably, the pixel space scrambling is specifically: constructing chaotic mapping based on an initial chaotic system, generating a chaotic sequence of pseudo random numbers, changing the spatial position of a target area pixel by performing primary matrix conversion of a limited step length on an image matrix, and changing an original area image into a disordered invisible new area image;
the chaotic encryption includes: reading in a sensitive area picture to be processed, entering a chaotic sequence through an encryption key, and adopting a chaotic system to design an encryption algorithm to realize the encryption purpose; and generating a chaotic key sequence related to the characteristics of the plaintext image by adopting a chaotic system and a logic diagram, replacing the pixel value of the target area, realizing the disturbance of the pixel position, and finally diffusing the image to obtain the target area encrypted image.
Preferably, the process of constructing the new dynamic chaotic encryption system is as follows: one chaotic map is used for controlling the parameters of the other chaotic system to form a new chaotic system, and the mathematical expression is as follows:
x n+1 ＝G(u n+1 ，x n )
u n+1 ＝H(y n+1 )
y n+1 ＝F(v，y n )
where F (x) represents the control chaotic map, G (x) represents the seed chaotic map, and H (x) represents the conversion relation function, which is used to map the output of F (x) into the parameter range of G (x).
Preferably, the image data stitching process includes:
and carrying out multi-channel decomposition on the encryption area image to obtain an encryption feature matrix of each channel of the encryption area image, obtaining a feature vector of each channel of the encryption area image, combining the feature vector with the unencrypted area feature matrix to generate an image matrix with the same size as the original image, and generating a target encryption image based on the combined encryption matrix for storage or transmission.
The system used in the method comprises:
the image automatic screening and matching module is used for screening and matching the input images according to conditions;
the image sensitive object recognition module is used for recognizing sensitive objects of image contents;
and the image dynamic chaotic encryption module is used for encrypting the image sensitive area.
The invention has the technical effects and advantages that: the method comprises the steps of carrying out structural feature extraction on an image based on an automatic identification image screening mechanism of an image quality screening and feature extraction method, and carrying out similarity matching on the extracted features to realize image screening of specific contents; through image target detection and feature comparison, the screened image is subjected to sensitive area identification, the sensitivity level is calculated according to sensitive objects contained in the image, and relevant sensitive areas meeting the conditions are segmented and extracted; the specific sensitive area is scrambled by using a pixel space scrambling method based on the chaotic sequence, then a new chaotic sequence is constructed by utilizing dynamic chaotic mapping, the scrambled image is diffused to realize encryption, and the encrypted image is obtained for storage or transmission by splicing the encrypted part of the sensitive area and the unencrypted part of the background area, so that the light-weight encryption is realized, the better safety performance is realized, and the coverage attack, the differential attack and the statistical attack of image data can be resisted.
Drawings
FIG. 1 is a schematic diagram of a system provided by the present invention;
FIG. 2 is a schematic diagram of an image screening process according to the present invention;
FIG. 3 is a schematic diagram of an image sensitive area identification process provided by the present invention;
FIG. 4 is a schematic diagram of an encryption flow of an image sensitive area provided by the invention;
fig. 5 is a gray scale frequency versus histogram of an image encryption color channel provided by the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention provides a lightweight dynamic image data encryption method and system as shown in figures 1-5, and the lightweight dynamic image data encryption method and system comprises the following steps:
step one: automatic screening of images: extracting structural features of the images based on an automatic image screening mechanism, and matching the similarity of the extracted features to realize the image screening of specific contents;
step two: sensitive object identification: image sensitive object recognition is carried out on the screened image through image target model detection and feature comparison, sensitivity is calculated according to the sensitive object contained in the image, and relevant sensitive areas meeting the conditions are extracted;
step three: dynamic chaotic encryption: scrambling a specific sensitive region by using a regional pixel space scrambling method based on a chaotic sequence, then constructing a new chaotic sequence by using dynamic chaotic encryption mapping, diffusing the scrambled image to realize dynamic chaotic encryption of the image, splicing an encrypted part of the sensitive region and an unencrypted part of a background region, and obtaining an encrypted image for storage or transmission.
Specifically, the method for automatically screening the images specifically comprises the following steps:
(1.1) image quality screening: filtering and denoising the original image, calculating pixel point edge gradients, accumulating and summing the pixel points, and normalizing the edge gradients to screen out clear images;
(1.2) similarity matching: carrying out structural feature extraction on the images meeting the quality requirements through a clustering and classifying algorithm, and screening out images matched with specific targets from the structural feature vectors based on a multi-metric similarity matching algorithm;
the sensitive object identification specifically comprises the following steps:
(2.1) sensitive target detection: identifying, detecting and classifying specific target objects contained in the image;
(2.2) sensitivity calculation: correspondingly calculating a sensitive target object according to a preset sensitivity value;
(2.3) sensitive region extraction: carrying out regional image segmentation extraction on a region which meets the sensitivity requirement of the target object;
the dynamic chaotic encryption comprises the following steps:
(3.1) area pixel scrambling: according to the initial chaotic system, the bytes of the initial image are subjected to preset mess processing, and the obtained array is subjected to dimension reduction by copying the RGB value of each adjacent pixel to obtain a preset mess bitmap.
(3.2) dynamic chaotic encryption: carrying out iterative diffusion on the bitmap pixels subjected to disorder processing according to a sequence of chaotic mapping by dynamically constructing a new chaotic system to obtain a ciphertext matrix;
(3.3) image data stitching: and splicing and integrating the mixed encrypted regional image ciphertext matrix and the unprocessed region of the original image to form the complete data of the target encrypted image.
Specifically, the process of the image quality screening specifically includes:
(1.11) performing image filtering noise reduction treatment, converting an original image into the filtering noise reduction treatment, filtering noise and smoothing the image, and removing irrelevant details in the image;
(1.12) calculating the edge gradients of the image in the neighborhood of each pixel point, respectively calculating the edge gradients of the neighborhood of each pixel point, carrying out weighted summation on the edge gradients of the neighborhood of each pixel point, calculating the edge gradient sum of each pixel point, accumulating the edge gradient sum of each pixel point, dividing the edge gradient sum by the number of the pixel points to obtain normalized edge gradients, and screening out a clear image according to the edge gradient value obtained after normalization.
Referring to fig. 2, the image filtering and noise reduction process specifically includes: the original image is converted to be filtered and noise reduced, the spatial proximity and pixel value similarity of the image are combined based on nonlinear bilateral filtering, meanwhile, the spatial proximity information and the color similarity information are considered, noise is filtered, the image is smoothed, meanwhile, edge preservation is achieved, and irrelevant details in the image are removed. And optimizing each weight calculated by the spatial proximity from each point to the center point, optimizing the weight to be the product of the weight calculated by the spatial proximity and the weight calculated by the pixel value similarity, and performing convolution operation on the optimized weight and the image.
The image edge gradient process specifically comprises the following steps: the computing process of the gradient value among the image edge pixel points is used for representing the edge information of the image, can reflect the changing speed of the image pixels and can indicate the definition degree of the image to a certain extent; respectively calculating edge gradients in the neighborhood of each pixel point for the pixel points of the image after the noise reduction treatment, accumulating the edge gradient sum of each pixel point, and dividing the sum by the number of the pixel points to obtain normalized edge gradients; and (3) screening out a clear image according to the edge gradient value obtained after normalization, wherein the gradient of the image f (x, y) at a certain point can be expressed as:
wherein G is x And G y Represented as a sequence of gradient chaos at a point.
In particular, the similarity matching includes feature clustering and feature classification,
the feature cluster specifically comprises the following contents:
(1.201) extracting image features of the image to obtain sample image features, and extracting structural features of the image features by utilizing a feature cluster model to obtain at least one image structural feature;
(1.202) clustering the image structural features based on the structural features to obtain at least one clustering result of the structural features;
(1.203) optimizing a preset feature clustering model according to a clustering result until the preset model converges to obtain a feature clustering model and the feature clustering model is used for screening structural features of a target image;
the feature classification specifically comprises:
(1.204) randomly and repeatedly selecting a plurality of image samples from the images, marking feature class labels as training samples, selecting an existing convolutional neural network model for parameter adjustment and optimization, and keeping the model structure and training data different between the convolutional neural network model and the network model for extracting the features;
(1.205) identifying the target image features by using a feature classification model to obtain a structural feature model and the category confidence coefficient thereof, and screening out the features as irrelevant samples if the confidence coefficient of a certain sample on the category to which the sample belongs is very low;
referring to fig. 1, the structural features of the images meeting the quality requirements are extracted through a clustering and classifying algorithm. Firstly, extracting image features of the image to obtain sample image features, extracting structural features of the image features by using a feature clustering model, clustering the image structural features, and adjusting parameters of a preset feature clustering model according to a clustering result until the preset model converges to obtain a feature clustering model which is used for screening the structural features of the target image. The classification method is characterized in that a plurality of image samples are selected from the images which can be randomly placed back, feature class labels are marked as training samples, an existing convolutional neural network model is selected for parameter adjustment and optimization, and feature classification models are utilized for identifying target image features to obtain a structural feature model and class confidence coefficients thereof.
The structural feature extraction comprises the following steps:
(1.21) defining structural information from the view of image composition as an attribute reflecting the structure of an object in a scene independent of brightness and contrast, taking a mean value as an estimation of the brightness, a standard deviation as an estimation of the contrast, and a covariance as a measure of the degree of structural similarity;
and (1.22) dividing the image into blocks by utilizing a sliding window, calculating based on pixels in the window in each step, and taking the average value as the structural similarity measure of the two images.
Referring to fig. 1, the structural feature vector is screened out to match with a specific target based on a multi-metric similarity matching algorithm. Structural similarity from the view of image composition, structural information is defined as the attribute of the object structure in the reflecting scene independent of brightness and contrast, the larger the value range [0,1], the smaller the image distortion is represented. Modeling is a combination of three different factors of brightness, contrast and structure, using the mean as an estimate of brightness, the standard deviation as an estimate of contrast, and the covariance as a measure of the degree of structural similarity. The distortion of the image varies in space, the image is segmented by a sliding window, the total number of the segmented blocks is N, and each step is calculated based on pixels in the window. Taking the influence of window shapes on the blocks into consideration, taking a Gaussian weighting function with standard deviation of 1.5 as a weighted window, and calculating the mean, variance and covariance of each window by adopting Gaussian weighting; and calculating the structural similarity of the corresponding blocks, and finally taking the average value as the structural similarity measure of the two images.
Specifically, the sensitive target detection specifically comprises the following steps:
according to a preset detection sensitive object list, automatically classifying and identifying target objects contained in the image content to obtain corresponding category information and position information;
extracting features according to the training images of the sensitive objects by using the target detection recognition, and using the obtained feature vectors for training a feature library for target image recognition to obtain a target detection model;
the sensitivity calculation method comprises the following steps:
calculating the sensitivity of the object in the current image containing the object according to the sensitive target detection identification information, wherein the sensitivity depends on the application scene of the image, the target preset value and the semantic relation among the objects in the image scene; the target preset value may employ an initial sensitivity of a predefined given target object in accordance with a general understanding of the image;
the sensitive area extraction process comprises the following steps:
and dividing the original image into a plurality of mutually disjoint areas according to the characteristics of gray scale, color, space texture, geometric shape and the like, including instance segmentation and scene segmentation, so as to obtain an image of the region of the target sensitive object.
The target object sensitivity calculation gives a calculation formula of image context sensitivity on the basis of experiments and analysis:
SD r ＝S r *W s +C r *W c +A r *W A +L r *W L ；
wherein W is s +W c +W A +W L =1, the region contour feature of the identified image content is S r Of the type characterised by C r The area ratio is A r The position distance is characterized by L r Each corresponding initial preset weight is W respectively S ，W C ，W A ，W L 。
Specifically, the pixel space scrambling specifically includes: constructing chaotic mapping based on an initial chaotic system, generating a chaotic sequence of pseudo random numbers, changing the spatial position of a target area pixel by performing primary matrix conversion of a limited step length on an image matrix, and changing an original area image into a disordered invisible new area image;
the chaotic encryption includes: reading in a sensitive area picture to be processed, entering a chaotic sequence through an encryption key, and adopting a chaotic system to design an encryption algorithm to realize the encryption purpose; and generating a chaotic key sequence related to the characteristics of the plaintext image by adopting a chaotic system and a logic diagram, replacing the pixel value of the target area, realizing the disturbance of the pixel position, and finally diffusing the image to obtain the target area encrypted image.
Specifically, the process of constructing the new dynamic chaotic encryption system is as follows: one chaotic map is used for controlling the parameters of the other chaotic system to form a new chaotic system, and the mathematical expression is as follows:
x n+1 ＝G(u n+1 ，x n )
u n+1 ＝H(y n+1 )
y n+1 ＝F(v，y n )
where F (x) represents the control chaotic map, G (x) represents the seed chaotic map, and H (x) represents the conversion relation function, which is used to map the output of F (x) into the parameter range of G (x).
Referring to fig. 4, it can be seen that:
regional pixel scrambling: based on the initial chaotic system, a chaotic map is constructed, a chaotic sequence of pseudo random numbers is generated, and the bytes of the initial image are subjected to preset mess processing. The order of the odd rows and columns of the initial image is reversed, then the RGB values of the rows and columns are rotated, and the obtained 3-dimensional array is transformed into a 2-dimensional array by copying the R value, G value and B value of each adjacent pixel, so as to obtain a preset disorder map with the size of mx 3N.
Dynamic chaotic encryption: and reading in the sensitive area picture to be processed, and carrying out iterative diffusion on the bit map pixels subjected to disorder processing according to a chaotic mapping sequence by dynamically constructing a new chaotic system so as to realize the encryption purpose. The dynamic construction process of the novel chaotic system comprises the steps of controlling parameters of the Tent chaotic system by using the Chirikov chaotic map to form the novel chaotic system, wherein the Chirikov chaotic system has the expression:
the equation of the Tent chaotic system is defined as:
calculating the sum of pixel values of the images by using the preset messy digital images with the size of M multiplied by 3N and recording the sum as s, and calculating the number k of initial iteration of the chaotic system by using the following formula, namely:
k＝s mod 10 3 +10 3
iteration M x 3N-1 times of the Chirikov chaotic system is carried out to obtain a pseudo-random number t1, and the following processing is carried out on the pseudo-random number t 1:
t＝mod((t 1 *10 3 )，2)+1
and (3) taking the t obtained in the step as a control parameter mu and iteration times k of the Tent chaotic system, carrying out iteration on the Tent chaotic system from the k times to obtain a chaotic sequence with the length of M multiplied by 3N, and sequencing. And (3) scrambling the pixels of the original image by using the index position sequence to obtain a scrambled image C'.
The chaotic mapping system is iterated by the input keys G0 and H0, and the chaotic sequences G and H with the length of M multiplied by 3N after iteration are intercepted. A new chaotic sequence F is constructed by two chaotic sequences according to the following formula, so that the new chaotic sequence F has good pseudo-randomness.
Generating a chaotic key sequence related to the characteristics of the plaintext image by adopting a chaotic system and a logic diagram, performing exclusive OR operation on the obtained chaotic sequence F and the scrambled image C', replacing the pixel value of the target area, realizing the scrambling of the pixel position, and finally diffusing the image to obtain the encrypted image of the target area:
C＝F⊕C’。
specifically, the image data stitching process includes:
and carrying out multi-channel decomposition on the encryption area image to obtain an encryption feature matrix of each channel of the encryption area image, obtaining a feature vector of each channel of the encryption area image, combining the feature vector with the unencrypted area feature matrix to generate an image matrix with the same size as the original image, and generating a target encryption image based on the combined encryption matrix for storage or transmission.
The system used in the method comprises:
the image automatic screening and matching module is used for screening and matching the input images according to conditions;
the image sensitive object recognition module is used for recognizing sensitive objects of image contents;
and the image dynamic chaotic encryption module is used for encrypting the image sensitive area.
Referring to fig. 5, the Lena image is taken as a sample image, the original image size is 256×256, and in the figures, (a), (B) and (c) are histograms of three channels of original pictures R, G and B, it can be seen that the distribution of pixels of the gray level histogram of the plaintext image is uneven, and the fluctuation is large; in the figures, (d), (e) and (f) are histograms of three channels of the encrypted image, the encrypted areas of the three figures are quite close, namely, the frequency of occurrence of each gray value is basically the same, the histogram distribution of pixels is more consistent, the information is converged, the correlation among pixels of the plaintext image is broken, and thus an attacker cannot acquire the related information of the sensitive area of the image. In addition, each channel of the encrypted picture has good randomness, can resist statistical attack, has correlation of horizontal, vertical and diagonal adjacent elements of each channel of the encrypted picture very close to 0, and greatly weakens the correlation of adjacent pixel points.
The Mean Square Error (MSE) characterizes the difference between the original image and the encrypted image, and when MSE is 0, it means that the two images are equal. Through analysis, when the value range of the key parameter enters a small interval, the picture can be decrypted, and the encrypted image can be effectively prevented from being cracked by using coverage attack. With the increase of randomness, the mean square error of the decrypted image is always increasing, and the increasing speed is in a slow trend. Meanwhile, the peak signal-to-noise ratio of the decrypted image is always reduced, and the reduction speed is in a slow trend.
The pixel change rate indicates that when any pixel value in the plaintext image has tiny change or the key has tiny change, the information of the ciphertext image can be greatly changed, and the closer to an ideal value, the better the sensitivity of the ciphertext to the plaintext is; the normalized average change intensity, when the values of the two indexes are closer to the ideal value, shows that the sensitivity to the subtle change of the plaintext image is higher, and the attack resistance is higher. The better pixel change rate and the normalized average change intensity are obtained through calculation, and the differential attack is better resisted.
Finally, it should be noted that: the foregoing description is only illustrative of the preferred embodiments of the present invention, and although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments described, or equivalents may be substituted for elements thereof, and any modifications, equivalents, improvements or changes may be made without departing from the spirit and principles of the present invention.
Claims (5)
1. A lightweight dynamic image data encryption method, comprising:
step one: automatic screening of images: extracting structural features of the images based on an automatic image screening mechanism, and matching the similarity of the extracted features to realize the image screening of specific contents;
step two: sensitive object identification: image sensitive object recognition is carried out on the screened image through image target model detection and feature comparison, sensitivity is calculated according to the sensitive object contained in the image, and relevant sensitive areas meeting the conditions are extracted;
step three: dynamic chaotic encryption: scrambling a specific sensitive region by using a regional pixel space scrambling method based on a chaotic sequence, then constructing a new chaotic sequence by using dynamic chaotic encryption mapping, diffusing the scrambled image to realize dynamic chaotic encryption of the image, splicing an encryption part of the sensitive region and an unencrypted part of a background region, and obtaining an encrypted image for storage or transmission;
the method for automatically screening the images in the first step specifically comprises the following steps:
(1.1) image quality screening: filtering and denoising the original image, calculating pixel point edge gradients, accumulating and summing the pixel points, and normalizing the edge gradients to screen out clear images;
(1.2) similarity matching: carrying out structural feature extraction on the images meeting the quality requirements through a clustering and classifying algorithm, and screening out images matched with specific targets from the structural feature vectors based on a multi-metric similarity matching algorithm;
wherein the similarity matching includes feature clustering and feature classification,
the feature cluster specifically comprises the following contents:
(1.201) extracting image features of an image to obtain sample image features, and extracting structural features of the image features by using a feature cluster model to obtain at least one image structural feature;
(1.202) clustering the image structural features based on the structural features to obtain at least one clustering result of the structural features;
(1.203) optimizing a preset feature clustering model according to a clustering result until the preset feature clustering model converges to obtain a feature clustering model and the feature clustering model is used for screening structural features of a target image;
the feature classification specifically comprises:
(1.204) randomly and repeatedly selecting a plurality of image samples from the images, marking feature class labels as training samples, selecting an existing convolutional neural network model for parameter adjustment and optimization, and keeping the model structure and training data different between the convolutional neural network model and the network model for extracting the features;
(1.205) identifying the target image features by using a feature classification model to obtain a structural feature model and the category confidence coefficient thereof, and screening out the features as irrelevant samples if the confidence coefficient of a certain sample on the category to which the certain sample belongs is low;
the structural feature extraction comprises the following steps:
(1.21) defining structural information from the view of image composition as an attribute reflecting the structure of an object in a scene independent of brightness and contrast, taking a mean value as an estimation of the brightness, a standard deviation as an estimation of the contrast, and a covariance as a measure of the degree of structural similarity;
(1.22) dividing the image into blocks by utilizing a sliding window, calculating based on pixels in the window in each step, and taking the average value as the structural similarity measure of the two images;
the sensitive object identification in the second step specifically comprises the following steps:
(2.1) sensitive target detection: identifying, detecting and classifying specific target objects contained in the image;
(2.2) sensitivity calculation: correspondingly calculating a sensitive target object according to a preset sensitivity value;
(2.3) sensitive region extraction: carrying out regional image segmentation extraction on a region which meets the sensitivity requirement of the target object;
the sensitive target detection comprises the following specific contents:
according to a preset detection sensitive object list, automatically classifying and identifying target objects contained in the image content to obtain corresponding category information and position information;
extracting features according to the training images of the sensitive objects by using the target detection recognition, and using the obtained feature vectors for training a feature library for target image recognition to obtain a target detection model;
the sensitivity calculation method comprises the following steps:
calculating the sensitivity of the object in the current image containing the object according to the sensitive target detection identification information, wherein the sensitivity depends on the application scene of the image, the target preset value and the semantic relation among the objects in the image scene; the target preset value adopts the initial sensitivity of a preset target object;
the sensitive area extraction process comprises the following steps:
dividing an original image into a plurality of mutually disjoint areas according to gray level, color, space texture and geometric shape characteristics, wherein the image comprises instance segmentation and scene segmentation, so as to obtain an image of the region of the target sensitive object;
the step three dynamic chaotic encryption comprises the following steps:
(3.1) area pixel scrambling: according to the initial chaotic system, carrying out preset mess processing on bytes of the initial image, and carrying out dimension reduction on the obtained array by copying RGB values of each adjacent pixel to obtain a preset mess bitmap;
(3.2) dynamic chaotic encryption: carrying out iterative diffusion on the bitmap pixels subjected to disorder processing according to a sequence of chaotic mapping by dynamically constructing a new chaotic system to obtain a ciphertext matrix;
(3.3) image data stitching: splicing and integrating the mixed encrypted regional image ciphertext matrix and an unprocessed region of the original image to form complete data of the target encrypted image;
the image data stitching process comprises the following steps:
and carrying out multi-channel decomposition on the encryption area image to obtain an encryption feature matrix of each channel of the encryption area image, obtaining a feature vector of each channel of the encryption area image, combining the feature vector with the unencrypted area feature matrix to generate an image matrix with the same size as the original image, and generating a target encryption image based on the combined encryption matrix for storage or transmission.
2. The method for encrypting lightweight dynamic image data according to claim 1, wherein said process of image quality screening is specifically as follows:
(1.11) performing image filtering noise reduction treatment, converting an original image into the filtering noise reduction treatment, filtering noise and smoothing the image, and removing irrelevant details in the image;
(1.12) calculating the edge gradients of the image in the neighborhood of each pixel point, respectively calculating the edge gradients of the neighborhood of each pixel point, carrying out weighted summation on the edge gradients of the neighborhood of each pixel point, calculating the edge gradient sum of each pixel point, accumulating the edge gradient sum of each pixel point, dividing the edge gradient sum by the number of the pixel points to obtain normalized edge gradients, and screening out a clear image according to the edge gradient value obtained after normalization.
3. The method for encrypting lightweight motion image data according to claim 1, wherein said spatial scrambling of regional pixels is specifically: constructing chaotic mapping based on an initial chaotic system, generating a chaotic sequence of pseudo random numbers, changing the spatial position of a target area pixel by performing primary matrix conversion of a limited step length on an image matrix, and changing an original area image into a disordered invisible new area image;
the dynamic chaotic encryption comprises: reading in a sensitive area picture to be processed, entering a chaotic sequence through an encryption key, and adopting a chaotic system to design an encryption algorithm to realize the encryption purpose; and generating a chaotic key sequence related to the characteristics of the plaintext image by adopting a chaotic system and a logic diagram, replacing the pixel value of the target area, realizing the disturbance of the pixel position, and finally diffusing the image to obtain the target area encrypted image.
4. A lightweight dynamic image data encryption method according to claim 3, wherein the dynamic chaotic encryption system is constructed by the following steps: one chaotic map is used for controlling the parameters of the other chaotic system to form a new chaotic system, and the mathematical expression is as follows:
；
wherein,representing the control of the chaotic map,representing the chaotic mapping of the seed,representing a conversion relation function, which is used to convertIs mapped to the output of (2)Is within the parameters of (a).
5. A system of a lightweight dynamic image data encryption method according to any one of claims 1 to 4, comprising:
the image automatic screening and matching module is used for screening and matching the input images according to conditions;
the image sensitive object recognition module is used for recognizing sensitive objects of image contents;
and the image dynamic chaotic encryption module is used for encrypting the image sensitive area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210948717.3A CN115378574B (en) | 2022-08-09 | 2022-08-09 | Lightweight dynamic image data encryption method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210948717.3A CN115378574B (en) | 2022-08-09 | 2022-08-09 | Lightweight dynamic image data encryption method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115378574A CN115378574A (en) | 2022-11-22 |
| CN115378574B true CN115378574B (en) | 2023-12-19 |
Family
ID=84063568
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210948717.3A Active CN115378574B (en) | 2022-08-09 | 2022-08-09 | Lightweight dynamic image data encryption method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115378574B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116226471B (en) * | 2023-05-10 | 2023-07-25 | 济宁蜗牛软件科技有限公司 | Data storage method for homeland resource planning |
| CN116431846B (en) * | 2023-06-14 | 2023-08-18 | 泰山学院 | Student sensitive information storage system and method |
| CN116451276B (en) * | 2023-06-15 | 2023-09-26 | 杭州海康威视数字技术股份有限公司 | Image processing method, device, equipment and system |
| CN116566585B (en) * | 2023-07-10 | 2023-09-05 | 中国科学院东北地理与农业生态研究所 | Informationized intelligent storage method |
| CN117278692B (en) * | 2023-11-16 | 2024-02-13 | 邦盛医疗装备（天津）股份有限公司 | Desensitization protection method for monitoring data of medical detection vehicle patients |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108650434A (en) * | 2018-05-08 | 2018-10-12 | 吉林大学 | A kind of method of image encryption |
| CN109977686A (en) * | 2019-04-01 | 2019-07-05 | 桂林电子科技大学 | A kind of image encryption method and image processing equipment based on Composite Chaotic System |
| CN111783116A (en) * | 2020-06-29 | 2020-10-16 | 南通职业大学 | Lightweight hyperchaotic rapid image encryption algorithm |
| WO2021227349A1 (en) * | 2020-05-11 | 2021-11-18 | 华南理工大学 | Front-end facial image encryption and recognition method for biometric privacy protection |
-
2022
- 2022-08-09 CN CN202210948717.3A patent/CN115378574B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108650434A (en) * | 2018-05-08 | 2018-10-12 | 吉林大学 | A kind of method of image encryption |
| CN109977686A (en) * | 2019-04-01 | 2019-07-05 | 桂林电子科技大学 | A kind of image encryption method and image processing equipment based on Composite Chaotic System |
| WO2021227349A1 (en) * | 2020-05-11 | 2021-11-18 | 华南理工大学 | Front-end facial image encryption and recognition method for biometric privacy protection |
| CN111783116A (en) * | 2020-06-29 | 2020-10-16 | 南通职业大学 | Lightweight hyperchaotic rapid image encryption algorithm |
Non-Patent Citations (2)
| Title |
|---|
| 刘进 ; 孙婧 ; 徐正全 ; 顾鑫 ; .基于目标的图像轻量级加密.华中科技大学学报(自然科学版).2011,(06),全文. * |
| 基于目标的图像轻量级加密;刘进;孙婧;徐正全;顾鑫;;华中科技大学学报(自然科学版)(06);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115378574A (en) | 2022-11-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115378574B (en) | Lightweight dynamic image data encryption method and system | |
| Al-Dmour et al. | A steganography embedding method based on edge identification and XOR coding | |
| Ubhi et al. | Neural style transfer for image within images and conditional GANs for destylization | |
| Sharma et al. | An enhanced Huffman-PSO based image optimization algorithm for image steganography | |
| Laishram et al. | A novel minimal distortion-based edge adaptive image steganography scheme using local complexity: (BEASS) | |
| Ulutas et al. | A new copy move forgery detection method resistant to object removal with uniform background forgery | |
| Hsu et al. | A high-capacity QRD-based blind color image watermarking algorithm incorporated with AI technologies | |
| CN111726472B (en) | Image anti-interference method based on encryption algorithm | |
| Meenakshi et al. | A hybrid matrix factorization technique to free the watermarking scheme from false positive and negative problems | |
| Zhang et al. | Exposing unseen GAN-generated image using unsupervised domain adaptation | |
| Liu et al. | Making DeepFakes more spurious: evading deep face forgery detection via trace removal attack | |
| Jana et al. | A new DCT based robust image watermarking scheme using cellular automata | |
| Zhang et al. | A local perturbation generation method for gan-generated face anti-forensics | |
| Cui et al. | Multitask identity-aware image steganography via minimax optimization | |
| Li et al. | High-capacity coverless image steganographic scheme based on image synthesis | |
| Dalal et al. | A secure video steganography scheme using DWT based on object tracking | |
| Mehrish et al. | Joint spatial and discrete cosine transform domain-based counter forensics for adaptive contrast enhancement | |
| Pilania et al. | An ROI-based robust video steganography technique using SVD in wavelet domain | |
| Nadimpalli et al. | Proactive deepfake detection using gan-based visible watermarking | |
| Alkhowaiter et al. | Evaluating perceptual hashing algorithms in detecting image manipulation over social media platforms | |
| Fadoua et al. | Medical video watermarking scheme for telemedicine applications | |
| Ouyang et al. | A semi-fragile reversible watermarking method based on qdft and tamper ranking | |
| Kumari et al. | Optimized dwt based digital image watermarking and extraction using rnn-lstm | |
| CN113191380B (en) | Image evidence obtaining method and system based on multi-view features | |
| CN111444985B (en) | Image matching method based on histogram matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |