A hierarchical self-encrypting surveillance video transmission system under network isolation conditions

Abstract

The present invention belongs to the field of video transmission technology, and discloses a hierarchical self-encrypting monitoring video transmission system under network isolation conditions. It includes a local intranet computer and video monitoring software installed on the intranet computer, a one-way network switch, a remote photoelectric monitoring device, a remote computer and a video transmission software installed on the remote computer; a deep learning method is used to analyze the content of the monitoring video screen, and a suitable video encryption method is dynamically selected to achieve dynamic adjustment between the encryption and decryption calculation amount, video bandwidth, and data security, so as to optimize the system performance. Specifically, it involves a method for determining the importance of video content, a method for encrypting high-importance videos, and a method for encrypting general-importance videos. The system can ensure the security of monitoring video data under the conditions of tight computing power resources of computer equipment and network isolation of the control center.

Description

Technical Field

[0001] The invention belongs to the technical field of video transmission, and relates to a hierarchical self-encrypting monitoring video transmission system under network isolation conditions. Background Art

[0002] The widespread application of high-definition video surveillance technology has saved a lot of human resources. However, considering the particularity of the application scenarios of the present invention and the sensitive information that may appear in the video data, the video data cannot be transmitted in plain text using the traditional frequency domain encoding method. Therefore, how to improve the security of surveillance video network transmission has become a problem that users have to consider.

[0003] Encrypting video data is a very effective technical measure to ensure the security of video data transmission. Since the encryption algorithm directly acts on the video data itself, as long as the encryption algorithm is not cracked, even if the video data is stolen, the valid information in the video will not be leaked.

[0004] Full-text encryption is a method of encrypting the code stream domain. It uses a general text encryption method, such as the Data Encryption Standard (DES) algorithm to encrypt the entire video stream. Since this method does not utilize the special structure of the video stream data, the amount of encryption and decryption calculations is quite large, and it is difficult to ensure the real-time performance of the surveillance video. In multimedia video, the motion vector data in the compressed domain often occupies more than half of the video bandwidth and has rich information. Therefore, there are also video encryption algorithms for motion vector data. Commonly used methods include encrypting the motion vector sign bit and scrambling the motion vector in the segment. Video encryption algorithms for motion vector data can generally achieve better performance in real time than encryption in the code stream domain. However, due to the characteristics of motion vector data, the encryption effect of a group of video sequences is not significant at the beginning. As the motion accumulates, the encryption effect continues to accumulate and the effect will become more and more significant. Therefore, this type of encryption method is not very suitable for video surveillance scenarios with high data security requirements. In view of the fact that the hybrid frequency domain coding technology based on discrete cosine transform (DCT) is widely used at present, this technology relies on DCT transform to effectively divide the pixel information in the video frame into low-frequency high-energy area and high-frequency low-energy area according to its color space. When the information of the high-frequency low-energy area is discarded, the impact on the overall pixel information of the video frame image is small. At the same time, the Zigzag arrangement and Huffman coding can greatly compress the obtained DCT coefficients. Therefore, in the encryption of image domain data, the DCT coefficients in the video data can be selected for encryption. There are two positions for encrypting the DCT coefficients. One is to encrypt after the DCT coefficients are generated, which can produce a good scrambling effect, but it will affect the compression coding based on correlation and seriously affect the compression efficiency. The other is to encrypt before encoding and transmission. This method has little impact on the bit rate, but the security is not as good as the former.

[0005] The video encryption algorithm (VEA-VideoEncryptionAlgorithm) changes the sign bits of the DC component and the AC component in the DCT coefficient by XORing the video data plaintext with the corresponding bits of the key. This method has a small amount of calculation and is very fast, but it can only obtain a simple encryption key and has slightly poor security.

[0006] The DCT coefficient layered encryption algorithm is to divide each macroblock into 16 blocks of 4×4, map the 16 DCT coefficients corresponding to the 16 blocks into a two-dimensional 16×16 coefficient matrix using the Zigzag function, and divide the matrix into 16 layers in row order, namely L i(0≤i<15), then selectively apply the layered scrambling algorithm to each layer. Generally speaking, the more layers of encryption, the better the security, but the higher the number of layers, the sparser the corresponding coefficients, so it is generally not necessary to encrypt all 16 layers, and encrypting the first 5 layers can achieve a better encryption effect. This method has a larger amount of calculation than the VEA algorithm, but it has better security.

[0007] Since the importance of surveillance video content changes according to the content in the picture, if a method can be found to automatically switch between different encryption methods according to the importance of the current surveillance video content to achieve dynamic adjustment among the encryption and decryption calculation amount, video bandwidth, and data security, the effect of leveraging the strengths of different encryption methods and avoiding their weaknesses can be achieved as much as possible.

[0008] In addition, in order to achieve the security of the entire process of monitoring video, the video decryption process should also be in a controlled environment. Therefore, network isolation technology can be used to allow video data to be transmitted into the intranet in one direction while limiting the outflow of intranet data. Video decryption can be completed in the intranet, avoiding the possibility of the decrypted video data and other intranet data being stolen by network attacks. Summary of the invention

[0009] The present invention provides a hierarchical self-encrypting surveillance video transmission system under network isolation conditions, which adopts a deep learning method to analyze the surveillance video content and dynamically selects a suitable video encryption method to achieve dynamic adjustment among the encryption and decryption calculation amount, video bandwidth, and data security, thereby optimizing system performance.

[0010] The technical solution adopted by the present invention is:

[0011] A hierarchical self-encrypting surveillance video transmission system under network isolation conditions, comprising a local intranet computer and video surveillance software installed on the intranet computer, a one-way network switch, a remote photoelectric monitoring device, a remote computer and video transmission software installed on the remote computer;

[0012] The video transmission software on the remote computer simultaneously pulls the video streams generated by the encoding of multiple remote optoelectronic monitoring devices, and uses deep learning methods to classify and identify the targets in the multi-channel monitoring screens one by one. If no important targets are found, the video is encrypted using the VEA encryption algorithm. If important targets are found, the discrete cosine transform coefficient layered encryption method is used to encrypt the video, and the encryption method information is written into the video stream. The video stream is then pushed to the local intranet computer through the UDP protocol through a one-way network switch, and the video stream data is processed using lossless encryption overlay technology;

[0013] After receiving the video stream, the video surveillance software of the local intranet computer parses the encryption method information and adopts the corresponding decryption method to decrypt the video data and then decode and play it.

[0014] Furthermore, a deep learning-based target detection model is used to classify and identify targets in multi-channel surveillance images to determine the importance of the current video clip.

[0015] Furthermore, when discrete cosine transform coefficient hierarchical encryption is selected, the current video line is further graded according to the importance of the target detection result. For video lines with an importance level less than a threshold, the first 5 layers of the DCT coefficient matrix after Zigzag function mapping are encrypted, and for video lines with an importance level greater than or equal to the threshold, the first 7 layers of the DCT coefficient matrix are encrypted.

[0016] Furthermore, when encrypting the video, the video transmission software writes the currently used encryption method and the key number information required for decryption into the auxiliary enhancement information of the video stream.

[0017] Furthermore, lossless encryption information superposition technology is used to convert the DCT coefficient matrices of several adjacent frames in the video image sequence of the video transmission software into an image complex domain model, which is then superimposed to generate a composite complex matrix and then converted into a standard video format for streaming.

[0018] The advantages of the present invention compared to the background technology are:

[0019] 1. Good security: The present invention encrypts the video data before network transmission, and automatically switches the encryption algorithm when different types of targets appear in the video screen, effectively improving the data security during the network transmission stage. In addition, the device for decrypting the video data is arranged in a network isolation environment, avoiding the leakage of the decrypted video data due to network attacks on the video decryption machine.

[0020] 2. Reduce the average computing power usage: The present invention can automatically switch the encryption algorithm when different types of targets appear in each video screen. Since there are no important targets in the screen most of the time in the video surveillance scenario, the present invention adopts an encryption method with slightly weaker security but less computing power usage when there are no important targets, and adopts an encryption method with strong security but more computing power usage when there are important targets. This can effectively reduce the average computing power usage while maintaining data security.

[0021] 3. Strong applicability: Even under poor network conditions and low bit rate video streaming, the integrity of the image frame information decoded by the video surveillance software can be guaranteed, effectively reducing the possibility of "flower screen" on the surveillance screen. At the same time, the image frame information is fully obfuscated and overlapped after encryption, further ensuring the security of video stream data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of the remote monitoring video transmission system in the present invention;

[0023] Figure 2 It is a video classification self-encryption decision flow chart of the present invention. DETAILED DESCRIPTION

[0024] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] like Figure 1 As shown, a hierarchical self-encrypting surveillance video transmission system under network isolation conditions includes a local intranet computer and video surveillance software installed on the intranet computer, a one-way network switch, a remote photoelectric monitoring device, a remote computer and video transmission software installed on the remote computer;

[0026] The video transmission software on the remote computer simultaneously pulls the video streams encoded by multiple remote optoelectronic monitoring devices, and uses a deep learning-based target detection model to classify and identify targets in multi-channel monitoring images to determine the importance of the current video clip. If no important target is found, the VEA encryption algorithm is used to encrypt the video. If an important target is found, the discrete cosine transform coefficient layered encryption method is used to encrypt the video, and the current encryption method and the key number information required for decryption are written to the auxiliary enhancement information of the video stream. The video stream is then pushed to the local intranet computer through a one-way network switch using the UDP protocol, and the video stream data is processed using lossless encryption overlay technology. The lossless encryption information overlay technology is used to convert the DCT coefficient matrices of adjacent frames of images in the video image sequence of the video transmission software into an image complex domain model, and the composite complex matrix is ​​superimposed to generate a standard video format for streaming.

[0027] After receiving the video stream, the video surveillance software of the local intranet computer parses the encryption method information and adopts the corresponding decryption method to decrypt the video data and then decode and play it.

[0028] Among them, when discrete cosine transform coefficient layered encryption is selected, the current video line is further graded according to the importance of the target detection result. For video lines with an importance level less than a threshold, the first 5 layers of the DCT coefficient matrix after Zigzag function mapping are encrypted, and for video lines with an importance level greater than or equal to the threshold, the first 7 layers of the DCT coefficient matrix are encrypted.

[0029] like Figure 2 As shown, the processing flow of this embodiment is:

[0030] 1. The video transmission software on the remote computer uses the RTSP protocol to simultaneously pull the video streams encoded by three remote optoelectronic monitoring devices;

[0031] 2. The video transmission software uses ffmpeg to decode the video stream data of the three lines, and then the deep learning module based on the YOLO-V4 model in the video transmission software classifies and identifies the targets in each video. If no important target (such as a ship or an airplane) is found or the confidence of the important target is lower than 0.5, the video transmission software uses the VEA method with less computational effort to encrypt the video. If an important target is found, the video transmission software uses the DCT coefficient layered encryption algorithm for encryption. For general important targets, the first 5 layers (i.e., L i (0≤i<5)), for very important targets, encrypt the first 7 layers (i.e. L i (0≤i<7)) to improve security;

[0032] 3. When the video transmission software uses the DCT coefficient layered encryption algorithm to encrypt the video screen with important targets, if the video screen does not show any important targets for two minutes, the video transmission software will automatically switch to the VEA encryption method for ordinary targets until the important targets appear again in the video screen;

[0033] 4. After encrypting the video stream data of the corresponding line, the video transmission software will first overlap the adjacent video frames of the same line and encryption method in the transmission video frame sequence through lossless encryption information superposition technology to ensure that the overlapped ciphertext information is transcoded into a standard video stream format;

[0034] 5. When encrypting a video, the video transmission software needs to write the encryption algorithm currently used and the key number required for decryption, etc., into the SEI (auxiliary enhancement information) of the video stream;

[0035] 6. The video transmission service on the remote computer uses the UDP protocol to push the video stream pulled from the remote monitoring device to the local intranet computer through a one-way network switch. The video data is received by the video monitoring software. The video monitoring software first parses the decoding information contained in the SEI, determines the decryption method and the corresponding key number to be used, and then extracts the key from the local code book for data decryption and subsequent decoding and display of the video.

Claims

1. A hierarchical self-encrypting surveillance video transmission system under network isolation conditions, characterized in that: Including local intranet computers and video monitoring software installed on the intranet computers, one-way network gateways, remote photoelectric monitoring equipment, remote computers and video transmission software installed on the remote computers; The video transmission software on the remote computer simultaneously pulls the video streams generated by the encoding of multiple remote optoelectronic monitoring devices, and uses deep learning methods to classify and identify the targets in the multi-channel monitoring screens one by one. If no important targets are found, the video is encrypted using the VEA encryption algorithm. If important targets are found, the discrete cosine transform coefficient layered encryption method is used to encrypt the video, and the encryption method information is written into the video stream. The video stream is then pushed to the local intranet computer through the UDP protocol through a one-way network switch, and the video stream data is processed using lossless encryption overlay technology; After receiving the video stream, the video surveillance software of the local intranet computer parses the encryption method information and adopts the corresponding decryption method to decrypt the video data and then decode and play it.

2. According to the hierarchical self-encrypting surveillance video transmission system under network isolation conditions of claim 1, it is characterized in that: A deep learning-based target detection model is used to classify and identify targets in multi-channel surveillance images to determine the importance of the current video clip.

3. The hierarchical self-encrypting surveillance video transmission system under network isolation conditions according to claim 1 is characterized in that: When discrete cosine transform coefficient hierarchical encryption is selected, the current video line is further graded according to the importance of the target detection result. For video lines with an importance level less than a threshold, the first five layers of the DCT coefficient matrix after Zigzag function mapping are encrypted. For video lines with an importance level greater than or equal to the threshold, the first seven layers of the DCT coefficient matrix are encrypted.

4. The hierarchical self-encrypting surveillance video transmission system under network isolation conditions according to claim 1 is characterized in that: When encrypting a video, the video transmission software writes the currently used encryption method and the key number information required for decryption into the auxiliary enhancement information of the video stream.

5. The hierarchical self-encrypting surveillance video transmission system under network isolation conditions according to claim 3 is characterized in that: By using lossless encryption information superposition technology, the DCT coefficient matrices of several adjacent frames in the video image sequence of the video transmission software are converted into image complex domain models, which are then superimposed to generate a composite complex matrix and then converted into a standard video format for streaming.

Images