CN112488930A - Sensitive image detection method and device based on anti-interference protection and electronic equipment - Google Patents
Sensitive image detection method and device based on anti-interference protection and electronic equipment Download PDFInfo
- Publication number
- CN112488930A CN112488930A CN202011156797.6A CN202011156797A CN112488930A CN 112488930 A CN112488930 A CN 112488930A CN 202011156797 A CN202011156797 A CN 202011156797A CN 112488930 A CN112488930 A CN 112488930A
- Authority
- CN
- China
- Prior art keywords
- image
- interference
- sensitive
- model
- neural network
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 46
- 238000000034 method Methods 0.000 claims abstract description 44
- 238000011896 sensitive detection Methods 0.000 claims abstract description 37
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims description 40
- 230000006870 function Effects 0.000 claims description 32
- 230000004913 activation Effects 0.000 claims description 28
- 238000010606 normalization Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 7
- 238000013527 convolutional neural network Methods 0.000 claims description 7
- 230000006872 improvement Effects 0.000 abstract description 4
- 230000008030 elimination Effects 0.000 description 12
- 238000003379 elimination reaction Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 8
- 238000013145 classification model Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
- G06F18/2193—Validation; Performance evaluation; Active pattern learning techniques based on specific statistical tests
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention provides a sensitive image detection method, a device and electronic equipment based on anti-interference protection, wherein the method comprises the following steps: inputting an image to be detected into a de-interference model, and outputting a restored image; the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image; inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image; the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label. The method, the device and the electronic equipment provided by the embodiment of the invention realize the improvement of the anti-interference capability, are suitable for adding different scenes with different interference in the sensitive image and save the computing resources.
Description
Technical Field
The invention relates to the technical field of sensitive image detection, in particular to a sensitive image detection method and device based on anti-interference protection and electronic equipment.
Background
Image classification is widely used in many fields of information security, and one typical application is sensitive image detection. Sensitive images refer to images that are spread across the internet with some malicious intent, such as weapons, riots, and malicious fraud. Sensitive image detection can identify whether content in the detected image comprises category sensitive information such as weapons, riots, malicious fraud and the like, and then filter out the sensitive images, so that the safety and privacy of network users are protected and network resources are saved. However, sensitive information distributors often create various disturbances to degrade the performance of image classification models, thereby hindering sensitive image detection. In particular, a new threat presents greater challenges for sensitive image detection, namely, immunity to interference. By adding a small but purposeful counter-disturbance to the regular image, a counter image is made that forces the classification model to output a wrong prediction. Fig. 1 is a schematic diagram of a countermeasure disturbance generation process provided in the prior art, and as shown in fig. 1, by adding a disturbance factor to an original image, a countermeasure sample image generated cannot identify weapon class sensitive information contained in the countermeasure sample image, cannot be classified into a weapon, and is wrongly classified into a cassette tape class. This imperceptible interference causes the sensitive image detection system to malfunction and keeps the sensitive image from being corrupted to propagate sensitive information. The interference resistance can even reduce the accuracy of image classification to 0%, which poses a serious threat to sensitive image detection systems and other security-sensitive artificial intelligence systems.
Designing an effective protection method to resist against interference threats becomes a key problem in the fields of information security and artificial intelligence. To protect against interference, most current research focuses on modifying the classification model to improve its robustness, such as generating more interfering images to enlarge the sample set of model training to retrain the model.
However, the current protection method has the following defects: the protection method for modifying the model only serves specific models and scenes, is poor in universality and requires a large amount of computing resources, and the defects cause that the method is difficult to effectively protect a sensitive image detection system.
Therefore, how to avoid the problems that the existing sensitive image detection method based on the anti-interference protection has poor anti-interference capability, the universality is poor due to the incapability of being applicable to various scenes, and the effective protection of a sensitive image detection system is difficult to form due to the need of a large amount of computing resources still remains to be solved by the technical personnel in the field.
Disclosure of Invention
The embodiment of the invention provides a sensitive image detection method and device based on anti-interference protection and electronic equipment, and aims to solve the problems that the existing sensitive image detection method based on anti-interference protection is poor in anti-interference capability, poor in universality caused by being incapable of being applied to various scenes and difficult to form effective protection on a sensitive image detection system due to the fact that a large amount of computing resources are needed.
In a first aspect, an embodiment of the present invention provides a method for detecting a sensitive image based on anti-interference protection, including:
inputting an image to be detected into a de-interference model, and outputting a restored image;
the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image;
inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image;
the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label.
In the method, the preset interference processing is to add Gaussian noise.
In the method, a neural network uses a CNN structure during the training of the interference removal model, and the neural network comprises N layers;
the layer 1 of the neural network sequentially comprises a convolution operation and a modified linear unit activation function, the layers 2 to N-1 of the neural network sequentially comprise a convolution operation, a batch normalization operation and a modified linear unit activation function, the layer N of the neural network comprises a convolution operation, and N is greater than 2 and is an integer.
In the method, the nth layer of the neural network sequentially comprises a convolution operation, a batch normalization operation and an output L of a modified linear unit activation functionnExpressed by the following formula:
wherein N is 2,3, …, N-1, kiIs a convolution kernel of the i-th layer of the neural network, Li-1For the output of the i-1 layer of the neural network, i ═ 2,3, …, n, BN (.) denotes a batch normalization function, and ReLU (·) denotes a modified linear element activation function.
In this method, the modified linear unit activation function relu (x) is max (0, x).
In a second aspect, an embodiment of the present invention provides a sensitive image detection apparatus based on anti-interference protection, including:
the restoration unit is used for inputting the image to be detected into the interference removal model and outputting a restored image; the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image;
the detection unit is used for inputting the restored image into a sensitive detection model and outputting a result whether the image to be detected belongs to a sensitive image; the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label.
In the device, the preset interference treatment is to add Gaussian noise.
In the device, a neural network uses a CNN structure during the training of the interference removal model, and the neural network comprises N layers;
the layer 1 of the neural network sequentially comprises a convolution operation and a modified linear unit activation function, the layers 2 to N-1 of the neural network sequentially comprise a convolution operation, a batch normalization operation and a modified linear unit activation function, the layer N of the neural network comprises a convolution operation, and N is greater than 2 and is an integer.
In a third aspect, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the steps of the method for detecting a sensitive image based on anti-interference protection as provided in the first aspect when executing the program.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the method for detecting a sensitive image based on anti-interference protection as provided in the first aspect.
According to the method, the device and the electronic equipment provided by the embodiment of the invention, the image to be detected is input into the interference elimination model, and the restored image is output; inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image; the interference elimination model is obtained after training based on a sample interference image and a corresponding original image label, the sample interference image is obtained after preset interference processing is carried out on the corresponding original image, and the sensitive detection model is obtained after training based on the sample interference-free image and a corresponding sensitive result label. Because the interference removing model is independently arranged, the interfered image can be restored to be close to the original image before interference, then the multiplexed image is used for carrying out sensitive detection, the negative influence of the interference on the sensitive detection model can be eliminated, the accuracy of the sensitive detection model is ensured, and the training of the interference removing model is only based on the simple sample interference image and the corresponding original image label, so that a large-scale sample set is not needed, and the calculation amount is reduced. Therefore, the method, the device and the electronic equipment provided by the embodiment of the invention realize the improvement of the anti-interference capability, are suitable for adding different interference scenes into the sensitive image and save the computing resources.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief description will be given below of the drawings required for the embodiments or the technical solutions in the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a schematic diagram of a counter interference generation process provided in the prior art;
fig. 2 is a schematic flowchart of a method for detecting a sensitive image based on anti-interference protection according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an interference cancellation model training network according to an embodiment of the present invention;
FIG. 4 is a schematic structural diagram of a sensitive image detection apparatus based on anti-interference protection according to an embodiment of the present invention;
fig. 5 is a schematic physical structure diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without inventive effort based on the embodiments of the present invention, are within the scope of the present invention.
The existing pollen particle detection based on the prior frame generally has the problems of low effectiveness, robustness and accuracy. In contrast, the embodiment of the invention provides a pollen detection method based on prior frame linear scaling. Fig. 2 is a schematic flow chart of a pollen detection method based on prior frame linear scaling according to an embodiment of the present invention, as shown in fig. 2, the method includes:
Specifically, firstly, the image to be detected is subjected to interference elimination processing, so that the image to be detected is restored to be an original image close to the original image before interference is added, and the interference elimination processing adopts an interference elimination model obtained through machine learning training. The interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image. To train the interference-free model, a training set of the model is first constructed, where the training set includes samples and corresponding labels, where an original color image is used as a label, interference is added to the color image through a preset interference process to form corresponding samples, and the preset interference process generally includes adding salt and pepper noise, adding triangular noise, adding white gaussian noise, and the like, and is not limited herein. And (3) constructing a training set, for example: and randomly extracting 1000 randomly selected color images, cutting each image into 4 pieces to form a training label set, adding Gaussian noise into the 4000 pieces to generate 4000 random interference pieces serving as a training sample set, and training a de-interference model based on the 4000 pieces and corresponding labels.
Specifically, the restored image is input into a sensitive detection model, and a result of whether the image to be detected belongs to the sensitive image is output, which is equivalent to that the sensitive detection model does not need to process interference problems possibly contained in the image, namely, the sensitive detection model is a traditional classification model, and whether an object in the image belongs to the categories of weapons, riots, malicious frauds and the like is identified to judge whether the image to be detected belongs to the sensitive image, and the sensitive detection model can be obtained only by training based on a normal non-interference image sample and a corresponding sensitive result label.
The method provided by the embodiment of the invention comprises the steps of inputting an image to be detected into an interference elimination model, and outputting a restored image; inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image; the interference elimination model is obtained after training based on a sample interference image and a corresponding original image label, the sample interference image is obtained after preset interference processing is carried out on the corresponding original image, and the sensitive detection model is obtained after training based on the sample interference-free image and a corresponding sensitive result label. Because the interference removing model is independently arranged, the interfered image can be restored to be close to the original image before interference, then the multiplexed image is used for carrying out sensitive detection, the negative influence of the interference on the sensitive detection model can be eliminated, the accuracy of the sensitive detection model is ensured, and the training of the interference removing model is only based on the simple sample interference image and the corresponding original image label, so that a large-scale sample set is not needed, and the calculation amount is reduced. Therefore, the method provided by the embodiment of the invention realizes the improvement of the anti-interference capability, is suitable for adding different interference scenes in the sensitive image and saves the computing resources.
Based on the above embodiment, in the method, the preset interference processing is to add gaussian noise.
Specifically, the preset interference processing is further limited to the most commonly used gaussian noise to reduce the complexity of the interference processing.
In accordance with any of the above embodiments, in the method,
the neural network uses a CNN structure during the training of the interference removal model, and comprises N layers;
the layer 1 of the neural network sequentially comprises a convolution operation and a modified linear unit activation function, the layers 2 to N-1 of the neural network sequentially comprise a convolution operation, a batch normalization operation and a modified linear unit activation function, the layer N of the neural network comprises a convolution operation, and N is greater than 2 and is an integer.
Specifically, the neural network uses a CNN structure in the interference elimination model training, and the neural network includes N convolutional layers, and each convolutional layer includes M convolutional filters, where M and N are both positive integers. Fig. 3 is a schematic structural diagram of an interference-free model training network according to an embodiment of the present invention, as shown in fig. 3, for a first layer of convolutional layers, which includes a convolution operation followed by a modified linear unit activation function, and convolutional layers in intermediate layers (i.e., layers 2 to N-1), where each layer includes a convolution operation, a batch normalization operation, and a modified linear unit activation function in sequence, and there is only one convolution operation in a last layer (i.e., layer N) of convolutional layers.
Based on any of the above embodiments, in the method, the nth layer of the neural network sequentially includes a convolution operation, a batch normalization operation, and an output L of the modified linear unit activation functionnExpressed by the following formula:
wherein N is 2,3, …, N-1, kiIs a convolution kernel of the i-th layer of the neural network, Li-1For the output of layer i-1 of the neural network, i-2, 3, …, n, BN (.) represents batch regressionThe normalization function, ReLU (), represents the modified linear unit activation function.
Specifically, the above equation specifically defines a calculation method for each convolutional layer in the neural network, and it is further explained herein that zero padding operation is required to be performed on input data of each convolutional layer to ensure that the data dimension size is not changed.
In any of the above embodiments, in the method, the modified linear unit activation function relu (x) is max (0, x).
In particular, the modified linear unit activation function relu (x) ═ max (0, x) is defined here even further, excluding some other forms of modified linear unit activation functions.
Based on any of the above embodiments, an embodiment of the invention provides a sensitive image detection apparatus based on anti-interference protection, fig. 4 is a schematic structural diagram of the sensitive image detection apparatus based on anti-interference protection provided in the embodiment of the invention, as shown in fig. 4, the apparatus includes a restoring unit 410 and a detecting unit 420, wherein,
the restoration unit 410 is configured to input the image to be detected into the interference-free model, and output a restored image; the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image;
the detection unit 420 is configured to input the restored image into a sensitive detection model, and output a result of whether the image to be detected belongs to a sensitive image; the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label.
According to the device provided by the embodiment of the invention, the image to be detected is input into the interference elimination model, and the restored image is output; inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image; the interference elimination model is obtained after training based on a sample interference image and a corresponding original image label, the sample interference image is obtained after preset interference processing is carried out on the corresponding original image, and the sensitive detection model is obtained after training based on the sample interference-free image and a corresponding sensitive result label. Because the interference removing model is independently arranged, the interfered image can be restored to be close to the original image before interference, then the multiplexed image is used for carrying out sensitive detection, the negative influence of the interference on the sensitive detection model can be eliminated, the accuracy of the sensitive detection model is ensured, and the training of the interference removing model is only based on the simple sample interference image and the corresponding original image label, so that a large-scale sample set is not needed, and the calculation amount is reduced. Therefore, the device provided by the embodiment of the invention realizes the improvement of the anti-interference capability, is suitable for adding different interference scenes in the sensitive image and saves the computing resources.
In accordance with any of the above embodiments, in the apparatus,
and the preset interference processing is to add Gaussian noise.
In accordance with any of the above embodiments, in the apparatus,
the neural network uses a CNN structure during the training of the interference removal model, and comprises N layers;
the layer 1 of the neural network sequentially comprises a convolution operation and a modified linear unit activation function, the layers 2 to N-1 of the neural network sequentially comprise a convolution operation, a batch normalization operation and a modified linear unit activation function, the layer N of the neural network comprises a convolution operation, and N is greater than 2 and is an integer.
In accordance with any of the above embodiments, in the apparatus,
the nth layer of the neural network sequentially comprises a convolution operation, a batch normalization operation and an output L of a modified linear unit activation functionnExpressed by the following formula:
wherein N is 2,3, …, N-1, kiIs a convolution kernel of the i-th layer of the neural network, Li-1Is layer i-1 of the neural networkI-2, 3, …, n, BN (.) denotes a batch normalization function, ReLU (.) denotes a modified linear unit activation function.
In accordance with any of the above embodiments, in the apparatus,
the modified linear unit activation function relu (x) max (0, x).
In order to verify that the sensitive image detection method based on anti-interference protection provided by the above embodiment can resist the security threat of interference to the sensitive image detection system, a plurality of sensitive images with weapon content are collected first, a deep neural network model Resnet50 is applied as a classification model, and an Accuracy (Accuracy) is used as a classification evaluation index. Firstly, 300 correctly classified sensitive images are selected for subsequent verification through a classification process, namely, the accuracy of the model classification original image is ensured to be 100%. Then 300 sensitive images against interference were generated using the representative Carlini and Wagner algorithm and the FGSM algorithm, respectively. The resulting countermeasure interference is ultimately classified using the Resnet50 model. In order to evaluate the effect of the protection method, 4 basic image interference removing algorithms (mean smoothing, median smoothing, gaussian smoothing and non-local average noise reduction) and the restoration unit of the invention are adopted to convert each 300 interfered images generated by the two algorithms, and then the converted images are input into the classification model Resnet50 again. Table 1 shows the evaluation results of the classification accuracy of the model before and after data conversion, and as shown in table 1, the anti-interference generated by the carini and Wagner algorithm has completely failed (all classified into non-weapon categories) only by using the conventional classification model. For the FGSM algorithm, the model correctly classified only 7.7% of the images. It follows that combating interference is indeed a serious threat to sensitive image detection systems. Through various methods, compared with 0% and 7.7% before conversion, the image processing method can improve the model classification accuracy to a certain extent, but the image classification accuracy converted by the recovery unit is 79.7% and 81.7% at most, and the threat of interference resistance is weakened to a certain extent. Therefore, the method provided by the invention is a feasible and effective way for protecting the sensitive image detection system and a plurality of artificial intelligence systems.
TABLE 1 evaluation results of model classification accuracy before and after data conversion
Fig. 5 is a schematic entity structure diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 5, the electronic device may include: a processor (processor)501, a communication Interface (Communications Interface)502, a memory (memory)503, and a communication bus 504, wherein the processor 501, the communication Interface 502, and the memory 503 are configured to communicate with each other via the communication bus 504. The processor 501 may call a computer program stored in the memory 503 and executable on the processor 501 to execute the method for detecting a sensitive image based on anti-interference protection provided by the above embodiments, for example, the method includes inputting an image to be detected into an interference elimination model, and outputting a restored image; the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image; inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image; the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label.
In addition, the logic instructions in the memory 503 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solutions of the embodiments of the present invention may be essentially implemented or make a contribution to the prior art, or may be implemented in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The embodiment of the present invention further provides a non-transitory computer readable storage medium, on which a computer program is stored, where the computer program is implemented by a processor to execute the method for detecting a sensitive image based on anti-interference protection provided in the foregoing embodiments, for example, the method includes inputting an image to be detected into an interference elimination model, and outputting a restored image; the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image; inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image; the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label.
The above-described system embodiments are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A sensitive image detection method based on anti-interference protection is characterized by comprising the following steps:
inputting an image to be detected into a de-interference model, and outputting a restored image;
the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image;
inputting the restored image into a sensitive detection model, and outputting a result whether the image to be detected belongs to a sensitive image;
the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label.
2. The method for detecting sensitive images based on anti-interference protection according to claim 1, wherein the preset interference treatment is adding Gaussian noise.
3. The method for detecting sensitive images based on anti-interference protection according to claim 1 or 2,
the neural network uses a CNN structure during the training of the interference removal model, and comprises N layers;
the layer 1 of the neural network sequentially comprises a convolution operation and a modified linear unit activation function, the layers 2 to N-1 of the neural network sequentially comprise a convolution operation, a batch normalization operation and a modified linear unit activation function, the layer N of the neural network comprises a convolution operation, and N is greater than 2 and is an integer.
4. The method for detecting sensitive image based on anti-interference protection according to claim 3, wherein the nth layer of the neural network comprises a convolution operation, a batch normalization operation and an output L of the modified linear unit activation function in sequencenExpressed by the following formula:
wherein N is 2,3, …, N-1, kiIs a convolution kernel of the i-th layer of the neural network, Li-1For the output of the i-1 layer of the neural network, i ═ 2,3, …, n, BN (.) denotes a batch normalization function, and ReLU (·) denotes a modified linear element activation function.
5. The method for detecting sensitive images based on anti-interference protection according to claim 4, wherein the modified linear unit activation function ReLU (x) max (0, x).
6. A sensitive image detection device based on anti-interference protection, comprising:
the restoration unit is used for inputting the image to be detected into the interference removal model and outputting a restored image; the interference removing model is obtained after training based on a sample interference image and a corresponding original image label, and the sample interference image is obtained after preset interference processing is carried out on the corresponding original image;
the detection unit is used for inputting the restored image into a sensitive detection model and outputting a result whether the image to be detected belongs to a sensitive image; the sensitive detection model is obtained by training based on the sample interference-free image and the corresponding sensitive result label.
7. The apparatus according to claim 6, wherein the predetermined interference processing is the addition of Gaussian noise.
8. The sensitive image detection device based on anti-interference protection according to claim 6 or 7,
the neural network uses a CNN structure during the training of the interference removal model, and comprises N layers;
the layer 1 of the neural network sequentially comprises a convolution operation and a modified linear unit activation function, the layers 2 to N-1 of the neural network sequentially comprise a convolution operation, a batch normalization operation and a modified linear unit activation function, the layer N of the neural network comprises a convolution operation, and N is greater than 2 and is an integer.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method for detecting sensitive images based on protection against interference according to any one of claims 1 to 5 when executing the program.
10. A non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the method for sensitive image detection based on protection against interference according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011156797.6A CN112488930A (en) | 2020-10-26 | 2020-10-26 | Sensitive image detection method and device based on anti-interference protection and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011156797.6A CN112488930A (en) | 2020-10-26 | 2020-10-26 | Sensitive image detection method and device based on anti-interference protection and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112488930A true CN112488930A (en) | 2021-03-12 |
Family
ID=74927237
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011156797.6A Pending CN112488930A (en) | 2020-10-26 | 2020-10-26 | Sensitive image detection method and device based on anti-interference protection and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112488930A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107122806A (en) * | 2017-05-16 | 2017-09-01 | 北京京东尚科信息技术有限公司 | A kind of nude picture detection method and device |
| CN107203765A (en) * | 2017-03-30 | 2017-09-26 | 腾讯科技（上海）有限公司 | Sensitive Image Detection Method and device |
| CN109145979A (en) * | 2018-08-15 | 2019-01-04 | 上海嵩恒网络科技股份有限公司 | sensitive image identification method and terminal system |
| CN109584221A (en) * | 2018-11-16 | 2019-04-05 | 聚时科技(上海)有限公司 | A kind of abnormal image detection method generating confrontation network based on supervised |
-
2020
- 2020-10-26 CN CN202011156797.6A patent/CN112488930A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107203765A (en) * | 2017-03-30 | 2017-09-26 | 腾讯科技（上海）有限公司 | Sensitive Image Detection Method and device |
| CN107122806A (en) * | 2017-05-16 | 2017-09-01 | 北京京东尚科信息技术有限公司 | A kind of nude picture detection method and device |
| CN109145979A (en) * | 2018-08-15 | 2019-01-04 | 上海嵩恒网络科技股份有限公司 | sensitive image identification method and terminal system |
| CN109584221A (en) * | 2018-11-16 | 2019-04-05 | 聚时科技(上海)有限公司 | A kind of abnormal image detection method generating confrontation network based on supervised |
Non-Patent Citations (1)
| Title |
|---|
| JIANGUO JIANG 等: "Restoration as a Defense Against Adversarial Perturbations for Spam Image Detection", 《ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING-ICANN 2019:IMAGE PROCESSING》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108549940B (en) | Intelligent defense algorithm recommendation method and system based on multiple counterexample attacks | |
| CN113554089B (en) | Image classification countermeasure sample defense method and system and data processing terminal | |
| CN109902617B (en) | Picture identification method and device, computer equipment and medium | |
| Maddireddy et al. | Enhancing Endpoint Security through Machine Learning and Artificial Intelligence Applications | |
| CN111260620B (en) | Image anomaly detection method and device and electronic equipment | |
| Vani | Towards efficient intrusion detection using deep learning techniques: a review | |
| Doan et al. | Defending backdoor attacks on vision transformer via patch processing | |
| Guan et al. | Badsam: Exploring security vulnerabilities of sam via backdoor attacks | |
| CN113114489B (en) | Network security situation assessment method, device, equipment and storage medium | |
| Chen et al. | Jujutsu: A two-stage defense against adversarial patch attacks on deep neural networks | |
| Li et al. | Temporal-distributed backdoor attack against video based action recognition | |
| Li et al. | Defending against insertion-based textual backdoor attacks via attribution | |
| Wang et al. | ADDITION: Detecting Adversarial Examples With Image-Dependent Noise Reduction | |
| Luo et al. | A novel defensive strategy for facial manipulation detection combining bilateral filtering and joint adversarial training | |
| CN112488930A (en) | Sensitive image detection method and device based on anti-interference protection and electronic equipment | |
| Ning et al. | Interpretation-Empowered Neural Cleanse for Backdoor Attacks | |
| Pal et al. | Towards understanding how self-training tolerates data backdoor poisoning | |
| CN115659387A (en) | Neural-channel-based user privacy protection method, electronic device and medium | |
| CN114930329A (en) | Method for training module and method for preventing capture of AI module | |
| Volodin et al. | Classification of attack mechanisms and research of protection methods for systems using machine learning and artificial intelligence algorithms | |
| Kimbrough et al. | Performance of gan-based denoising and restoration techniques for adversarial face images | |
| Gu | Watermark Removal Scheme Based on Neural Network Model Pruning | |
| Khan et al. | Adversarially Robust Deepfake Detection via Adversarial Feature Similarity Learning | |
| Yılmaz | Malware classification with using deep learning | |
| Hu et al. | A fast and effective image steganalysis model based on convolutional neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210312 |