A method for protecting image data based on compressed sensing

Abstract

The present application relates to a kind of protection methods of image data based on compressed sensing, comprising the steps of: collecting image data, the compressed sensing is carried out to the image data, and compressed picture data is obtained;Select encryption bit, the compressed image data is locally encrypted using AES-CTR encryption mode, and encrypted image data is obtained;Provide authentication information, use histogram to hide the authentication information in the encrypted picture data, and the encrypted image data hidden with the authentication information is transmitted to storage device;Extract the authentication information on the storage device, and the encrypted image data after verification is decrypted and decompressed, and original image data is obtained.The protection method of image data based on compressed sensing of the present application can effectively prevent the occurrence of data tampering and theft, ensure the confidentiality and integrity of image data during transmission, and also ensure the real-time of image data during transmission.

Description

Technical Field

[0001] This invention relates to the field of data protection, and more particularly to a method for protecting image data based on compressed sensing. Background Technology

[0002] A drone is an unmanned aircraft controlled by radio equipment or pre-installed control programs. The world's first drone was developed in the 1920s to meet the needs of warfare. Initially used only as non-recoverable "bombs," drones have seen continuous improvements in performance with technological advancements, finding widespread application in both military and civilian fields. Military drones are used for intelligence gathering, electronic warfare, and aerial reconnaissance, while civilian drones are used for urban mapping, pesticide spraying, and power line inspection.

[0003] During the data acquisition and transmission process of drones, the quality and efficiency of image data acquisition directly determine the overall performance of the entire system. However, due to limited computing power and storage space, these images are typically sensed and compressed by wireless image sensors before being sent to ground storage devices for further processing. This approach presents security challenges. Drones usually transmit information to ground storage devices via wireless networks; if this network is not encrypted, attackers can easily eavesdrop on any transmitted content. Therefore, it is crucial to strengthen the confidentiality and integrity of data during drone data acquisition and transmission. A common solution is to follow a compression-then-encryption approach, compressing images before encryption to save bandwidth. However, this method has a significant drawback: since both compression and encryption are computationally intensive, this results in a huge computational load and noticeable latency.

[0004] In addition to data confidentiality, data authenticity also needs to be considered. If drone data is maliciously intercepted and tampered with by attackers during transmission, the obtained information will be altered, posing a fatal threat to subsequent work. Therefore, verification information needs to be added during data transmission. Thus, to maintain data authenticity in adversarial environments, it is urgent to develop lightweight encryption and verification algorithms within limited resources and time constraints.

[0005] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0006] In view of the shortcomings of the prior art, the purpose of this application is to provide a method for protecting image data based on compressed sensing, which aims to solve the technical problems of large computational load and low security against illegal attacks in the transmission of image data from existing UAVs.

[0007] A method for protecting image data based on compressed sensing, comprising:

[0008] Image data is acquired, and the image data is compressed and sensed to obtain compressed image data;

[0009] Select encryption bits and use AES-CTR encryption mode to partially encrypt the compressed image data to obtain encrypted image data;

[0010] Provide verification information, use a histogram to hide the verification information in the encrypted image data, and transmit the encrypted image data containing the hidden verification information to a storage device;

[0011] The verification information is extracted from the storage device, and the verified encrypted image data is decrypted and decompressed to obtain the original image data.

[0012] In the above implementation, encrypted image data is first obtained through compression sensing and local encryption. Then, hidden verification information is inserted into the encrypted image data. The encrypted image data containing the hidden verification information is transmitted to a storage device, which can be a server. This ensures the stability of the image data during transmission and effectively prevents data tampering and theft due to external attacks. Furthermore, in this embodiment, when decrypting and decompressing the encrypted image data containing the hidden verification information on the storage device, verification is required first. After successful verification, the original image data is obtained through decryption and decompression. Before decryption, verification is performed to check whether the transmitted image data has been tampered with. If the image data has been tampered with, the decryption process is terminated directly. If the image data has not been tampered with, decryption and decompression are performed. This simplifies the process compared to existing technologies.

[0013] Optionally, the steps of retrieving the verification information from the storage device, decrypting and decompressing the verified encrypted image data to obtain the original image data include:

[0014] Extract the verification information. If the verification information has been tampered with, discard the encrypted image data directly. If the verification information has not been tampered with, decrypt and decompress the encrypted image data to obtain the original image data.

[0015] In the above implementation, considering that the image data may be maliciously intercepted or tampered with by attackers during transmission, if the verification information is found to have been tampered with when extracting the verification information, the encrypted image data containing the verification information is directly discarded. If the verification information is correct, the encrypted image data is decrypted and decompressed to obtain the original image data. By adding a step to verify the verification information, the validity of the transmitted encrypted image data is ensured, the decryption and decompression of invalid data is avoided, and the process is simplified.

[0016] Optionally, the step of performing compressed sensing on the image data to obtain compressed image data includes:

[0017] Let the image data be x, select an m×n (m<<n) Gaussian observation matrix φ, and use the formula y=φx+e=φψs+e=As+e to obtain the observation vector y corresponding to the image data x, where A=φψ is the holographic dictionary and e represents sampling noise or quantization error;

[0018] Provide a z (z>0, z∈N) * A quantizer Q is used to compress the image data x, resulting in compressed image data Q'.

[0019] Optionally, the Gaussian observation matrix φ is composed of a standard Gaussian distribution and needs to satisfy the constraint of isometry.

[0020] Optionally, the step of selecting encryption bits and using AES-CTR encryption mode to partially encrypt the compressed image data to obtain encrypted image data includes:

[0021] Select a stream cipher and use the AES-CTR encryption mode to obtain the corresponding index set, which is represented as: in,

[0022] Repeat the steps described above for obtaining the index set using the selected stream cipher to obtain [pm]-bit random numbers b1, b2, ..., b [pm] ;

[0023] Use the sign bit of index Q1 (i∈G) to set the random numbers b1, b2, ..., b[ pm Perform an XOR operation to obtain encrypted image data.

[0024] Optionally, the step of providing verification information and hiding the verification information in the encrypted image data using a histogram includes:

[0025] The verification information is represented as a vector w;

[0026] The vector w is embedded into the encrypted image data using a histogram, and the encrypted observations are updated accordingly.

[0027] Optionally, the step of extracting verification information and decrypting and decompressing the verified encrypted image data to obtain the original image data includes:

[0028] Extract the embedded hidden vector w, and obtain an approximate value c' of the encrypted observation vector through the hidden vector w;

[0029] Decrypt the sign bit of c' to recover the compressed image data Q';

[0030] The observation vector y' is approximated by inverse quantization encoding of the compressed image data Q'.

[0031] The original image data is obtained through CS reconstruction.

[0032] Optionally, the key used to decrypt the sign bit of c' is the stream cipher used during local encryption.

[0033] Optionally, the storage device is a server.

[0034] Optionally, the verification information includes ID information, location information, image acquisition time, and CRC64 encoding of the observation vector. Attached Figure Description

[0035] Figure 1 This is a flowchart of an image data protection method according to the present invention; Detailed Implementation

[0036] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings. Preferred embodiments of this application are shown in the drawings. However, this application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0038] In existing UAV image data acquisition and transmission processes, a common approach is to compress the data before encrypting it, thus saving bandwidth. However, this method results in significant computational overhead and noticeable latency. Furthermore, to maintain data confidentiality, traditional methods use message authentication codes based on cryptographic hash functions or block cipher algorithms to verify messages. However, research combining authenticity and confidentiality in CS (compressed sensing)-based security systems is almost nonexistent. This embodiment develops a method for protecting image data based on compressed sensing, such as... Figure 1 As shown, the method includes the following steps:

[0039] S10. Acquire image data, and perform compressed sensing on the image data to obtain compressed image data;

[0040] S20. Select encryption bits and use AES-CTR encryption mode to partially encrypt the compressed image data to obtain encrypted image data;

[0041] S30. Provide verification information, use a histogram to hide the verification information in the encrypted image data, and transmit the encrypted image data containing the hidden verification information to a storage device;

[0042] S40. Extract the verification information from the storage device, decrypt and decompress the verified encrypted image data to obtain the original image data.

[0043] Compressed sensing, also known as compressed sampling or compressed sensing, utilizes the sparsity of signals to obtain discrete samples of the signal using random sampling at a sampling rate much smaller than Nyquist. Subsequently, the signal is perfectly reconstructed using a nonlinear reconstruction algorithm. In this embodiment, encrypted image data is first obtained through compression sensing and local encryption. Then, hidden verification information is inserted into the encrypted image data. The encrypted image data containing the verification information is then transmitted to a storage device, which can be a server. This effectively prevents data tampering and theft due to external attacks, ensuring the confidentiality and integrity of the image data during transmission. Furthermore, considering the limitations of storage devices and computing power, this embodiment adopts a compression sensing-based image data protection method, ensuring real-time performance during transmission. In this embodiment, when decrypting and decompressing the encrypted image data containing the verification information on the storage device, verification is required first. After successful verification, the original image data is obtained through decryption and decompression. Before decryption, verification is performed to check whether the transmitted image data has been tampered with. If the image data has been tampered with, the decryption process is terminated directly. If the image data has not been tampered with, decryption and decompression are performed. This simplifies the process compared to existing technologies.

[0044] The image data protection method developed in this embodiment can ensure the data security of the drone during the data transmission process after data collection, enabling it to resist illegal attacks. Moreover, the computational load during transmission is small and the transmission process is simple.

[0045] In some implementations, the steps of retrieving verification information from the storage device, decrypting and decompressing the verified encrypted image data to obtain the original image data include:

[0046] Extract the verification information. If the verification information has been tampered with, discard the encrypted image data directly. If the verification information has not been tampered with, decrypt and decompress the encrypted image data to obtain the original image data.

[0047] In this embodiment, considering that image data may be maliciously intercepted or tampered with by attackers during transmission, if the verification information is found to have been tampered with when extracting verification information, the encrypted image data containing the verification information is directly discarded. If the verification information is correct, the encrypted image data is decrypted and decompressed to obtain the original image data. By adding a step to verify the verification information, the validity of the transmitted encrypted image data is ensured, the decryption and decompression of invalid data is avoided, and the process is simplified.

[0048] In some implementations, the step of compressing the image data to obtain compressed image data includes:

[0049] S101. Set the image data as x, select an m×n (m<<n) Gaussian observation matrix φ, and obtain the observation vector y corresponding to the image data x using the formula y=φx+e=φψs+e=As+e, where A=φψ is the holographic dictionary and e represents sampling noise or quantization error.

[0050] S102, Provide a z (z>0, z∈N) * A quantizer Q with 10 bits is used to compress the image data x to obtain compressed image data.

[0051] In this embodiment, the image captured by the drone is the original image. Represent the n-dimensional original image data as x. If there exists a ψ such that x = ψs, then the image data x is called k-sparse, where s is a vector, and the l0 norm of s satisfies ||s||0 = k, and k << n. The Gaussian observation matrix φ is composed of standard Gaussian distributions, generally composed of I.I.D (independent and identically distributed) standard Gaussian distributions, and the Gaussian observation matrix φ needs to satisfy the restricted isometry property. In this embodiment, by selecting a Gaussian observation matrix φ of m×n (m << n), then, according to the Gaussian observation matrix φ, using the formula y = φx + e = φψs + e = As + e, the observation vector y corresponding to the image data x can be obtained, where A = φψ is the holographic dictionary, and e represents sampling noise or quantization error.

[0052] Further, after obtaining the observation vector y corresponding to the image data x, a quantizer Q is used to compress the image data x. Quantization is the process of using the quantizer Q to compress the picture signal into a specified range so that the amplitude of the picture signal is represented by a digital code with a finite number of levels. The quantizer Q used in this embodiment is a uniform scalar quantizer with z bits (z > 0, z ∈ N * )), where Q' = Q(y) represents the quantized sample, and Q' is the compressed image data.

[0053] In some embodiments, the steps of selecting encryption bits and locally encrypting the compressed image data using the AES-CTR encryption mode to obtain encrypted image data include:

[0054] S201. Select a stream cipher and use the AES-CTR encryption mode to obtain a corresponding index set, which is represented as: where,

[0055] S202. Repeat the step of selecting the stream cipher to obtain the index set to obtain [pm]-bit random numbers b1, b2,....., b [pm ;

[0056] S203. Use the sign bits of the index items Q1 (i ∈ G) to perform exclusive OR on the random numbers b1, b2,....., b [pm to obtain the encrypted image data.

[0057] In this embodiment, first, generate the encryption bit index: Select a stream cipher to obtain an index set represented as , where, p is a parameter controlling the proportion of entries to be encrypted. p determines the decoding quality for unauthorized users, who are equivalent to malicious attackers. Then, the steps for generating the encryption bit index are repeated to obtain random numbers b1, b2, ..., b of [pm]-bit. [pm] The sign bit of index Q1 (i∈G) is used to set the random numbers b1, b2, ..., b [pm] Performing an XOR operation, a mathematical operator applicable to logical operations, yields encrypted image data and the corresponding observation vector. The encryption process can be described as follows:

[0058]

[0059] k is i = g k

[0060] The value at that time, where key is the selected stream cipher.

[0061] Furthermore, in some embodiments, the step of providing verification information and hiding the verification information in the encrypted image data using a histogram includes:

[0062] S301. Represent the verification information as a vector w;

[0063] S302. Embed the vector w into the encrypted image data using a histogram, and update the encrypted observation value to...

[0064] In this embodiment, the verification information includes ID information, location information, image acquisition time, and CRC64 encoding of the observation vector. The verification information is represented as a vector w. Subsequently, the vector w is embedded into the encrypted image data, and its corresponding encrypted observation value is updated. in,

[0065]

[0066] Further, the steps of extracting verification information, decrypting and decompressing the verified encrypted image data to obtain the original image data include:

[0067] S401. Extract the embedded hidden vector w, and obtain an approximate value c' of the encrypted observation vector through the hidden vector w;

[0068] S402. Decrypt the sign bit of c' to recover the compressed image data Q';

[0069] S403. The observation vector y' is approximated by inverse quantization encoding of the compressed image data Q'.

[0070] S404. Obtain the original image data through compressed sensing reconstruction.

[0071] In this embodiment, firstly, the embedded hidden vector w is extracted, if Then w = 0, if Then w = 1. Subsequently, by extracting information from the hidden vector w, an approximate value c' of the selected encrypted observation vector is obtained, where...

[0072]

[0073] When the approximate value c' of the observation vector is available, a verified user can recover the compressed image data Q' by decrypting the sign bit of the approximate value c' of the observation vector under the same key, and then approximate the observation vector y' by dequantizing the compressed image data Q'. Since the formula y = Φx + e = ΦΨs + e = As + e has infinite solutions, in order to accurately recover the original image data x, in this embodiment, based on the set min||Ψ -1 x||1 and ||y'-Ax||2≤ε e Given the constraints, we obtain s. Then, using the formula x = Ψs, we reconstruct the original image data x through compressed sensing (CS).

[0074] This invention develops a method for protecting image data based on compressed sensing, which effectively prevents data tampering and theft caused by external attacks, ensuring the confidentiality and integrity of the image data during transmission. Considering the limitations of storage devices and computing power, the compressed sensing-based image data protection method guarantees real-time performance during transmission. Furthermore, by inserting verification information, verification is performed before decryption to check whether the transmitted image data has been tampered with. If the image data has been tampered with, the decryption process is terminated directly; if the image data has not been tampered with, decryption and decompression proceed, simplifying the process compared to existing technologies.

[0075] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for protecting image data based on compressed sensing, characterized by, The method comprises the following steps: The method comprises: setting the image data as x, selecting a Gaussian observation matrix φ of m*n (m << n), the Gaussian observation matrix φ is composed of standard Gaussian distribution and needs to satisfy the constraint equidistance; obtaining the observation vector y corresponding to the image data x according to the Gaussian observation matrix φ, wherein A=φψ is a holographic dictionary, e represents sampling noise or quantization error; providing a quantizer Q of z (z>0, z∈N * ) bits, and compressing the picture data x by using the quantizer Q to obtain compressed picture data Q'; Selecting encryption bits, using AES-CTR encryption mode to locally encrypt the compressed image data to obtain encrypted image data, comprising: selecting a stream cipher, using AES-CTR encryption mode to obtain a corresponding index set, the index set is represented as: Wherein, Is a control parameter of the proportion of items that need to be encrypted, which determines the decoding quality of unauthorized users; repeating the step of selecting a stream cipher to obtain an index set to obtain [pm]-bit random numbers b1, b2,..., b[ pm ]; Using the sign bit of the index item Q1(i∈G) to XOR the random numbers b1, b2,..., b [pm] Obtain encrypted image data; the encryption process is described as: where k is the value of i = g k when key is the selected stream cipher; Providing authentication information, hiding the authentication information in encrypted picture data using a histogram, and transmitting the encrypted image data having the authentication information hidden therein to a storage device, includes representing the authentication information as a vector w, embedding the vector w in the encrypted image data using a histogram, and updating encrypted observations as wherein, Extracting the authentication information on the storage device, decrypting and decompressing the encrypted image data after authentication to obtain the original image data, comprising: extracting the embedded hidden vector w, if w = 0, if w = 1; through the information of the extracted hidden vector w, the approximate value c' of the selected encrypted observation vector is obtained, wherein, When the approximation c' of the observation vector is available, the verified user can recover the compressed image data Q' by decrypting the sign bits of the approximation c' of the observation vector under the same key, and get the observation vector y' by inverse quantization coding the compressed image data Q'; according to the set min||Ψ -1 x||1 and ||y'-Ax||2≤ε e The constraint condition is obtained s, and then the original image data x is obtained by using the formula x=Ψs through compressed sensing reconstruction.

2. The method of claim 1, wherein, The step of extracting the verification information, and decrypting and decompressing the encrypted image data after verification to obtain the original image data comprises the following steps: The verification information is extracted, if the verification information is tampered, the encrypted image data is directly discarded, if the verification information is not tampered, the encrypted image data is decrypted and decompressed to obtain the original image data. 3.The compression sensing based image data protection method of claim 1, wherein, The key used for decrypting the sign bit of c' is the stream cipher used in local encryption.

4. The method of claim 1, wherein, The storage device is a server.

5. The method of claim 1, wherein, The verification information comprises ID information, position information, picture collection time and CRC64 encoding of the observation vector.

Images