A secure communication method based on double-end deployment of multi-layer time-delay reserve pool calculation

Abstract

The present invention discloses a confidential communication method based on double-end deployment of multi-layer time delay reserve pool calculation, including: deploying a time delay reserve pool calculation model composed of multiple series time delay reserve pools to a receiving end and a sending end at the same time, using a training set L(t) to train the time delay reserve pool calculation models of the receiving end and the sending end at the same time; the chaotic system of the sending end continues to generate new chaotic signals, retaining Ux(t) as a driving set to drive the time delay reserve pool calculation model of the sending end to obtain a chaotic sequence Uz(t) for encryption, and sending Ux(t) to the time delay reserve pool calculation model of the receiving end to synchronously generate a chaotic sequence for decryption consistent with Uz(t); the sending end uses the encryption chaotic sequence to encrypt the communication data to be transmitted; the receiving end uses the decryption chaotic sequence to decrypt the received encrypted communication data. The present invention can realize high-quality chaotic synchronization between the transmitting end and the receiving end, thereby stably demodulating the transmitted useful information.

Description

Technical Field

[0001] The present invention relates to the cross-technical field of machine learning and chaotic secure communication, and in particular to a secure communication method based on double-end deployed multi-layer time-delay reserve pool calculation. Background Art

[0002] The core of chaotic secure communication lies in the realization of chaotic synchronization between the encryption system and the decryption system. In the traditional mechanism, the transmitter and the receiver generally need to configure the chaotic dynamic system separately, and achieve chaotic synchronization by designing the coupling mechanism, so as to demodulate the useful information transmitted. However, the traditional mechanism has the following disadvantages: 1. The selection of encryption system and confidentiality system is not flexible. Once implemented in hardware, it is difficult to replace it at low cost; 2. The implementation of the coupling mechanism is relatively complicated and needs to be carefully designed for different encryption / decryption mechanisms. Time-delay reservoir (TDR) computing, as a new machine learning algorithm, can be used for the synchronization of chaotic dynamic systems. By learning the time evolution data of a certain dimension of the dynamic system, the time evolution data of another dimension of the dynamic system can be predicted (synchronized).

[0003] The invention with publication number CN113556222A discloses a method for implementing secure communication based on multiple time-delay reserve pool calculations in series, and the invention with publication number CN112910625A discloses a method and system for secure communication based on machine learning reserve pool calculations. Both of them realize secure communication by deploying a time-delay reserve pool calculation structure at the receiving end. At present, there are no research results on the simultaneous deployment of a multi-layer structure of time-delay reserve pool calculations at both ends. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a confidential communication method based on double-end deployment of multi-layer time delay reserve pool calculation, wherein a multi-layer structure based on time delay reserve pool calculation is simultaneously deployed at the confidential communication transmitting / receiving end, and a novel chaotic synchronization implementation mechanism is implemented based on FPGA design, thereby performing confidential communication of various types of information.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The embodiment of the present invention proposes a secure communication method based on dual-end deployment of multi-layer time delay reserve pool calculation, and the secure communication method includes the following steps:

[0007] S1, using the chaotic system at the sending end to generate a multidimensional chaotic sequence, generating a training set L(t) containing three dimensions Lx(t), Ly(t) and Lz(t), and sending the training set L(t) to the receiving end;

[0008] S2, deploying a time delay reserve pool calculation model consisting of multiple series-connected time delay reserve pools to the receiving end and the sending end at the same time, and using the training set L(t) to train the time delay reserve pool calculation models of the receiving end and the sending end at the same time;

[0009] S3, the chaotic system at the sending end continues to generate new chaotic signals, and the generated new chaotic signals are recorded as the encryption set U(t). U(t) is the continuation of L(x) and contains three dimensions Ux(t), Uy(t) and Uz(t); Ux(t) is retained as the driving set to drive the time-delay reserve pool calculation model at the sending end to obtain the encryption chaotic sequence Uz(t) as the encryption set, and at the same time, Ux(t) is sent to the time-delay reserve pool calculation model at the receiving end to synchronously generate the decryption chaotic sequence U'z(t) consistent with Uz(t) as the solution set;

[0010] S4, the sending end uses the encryption chaotic sequence Uz(t) to encrypt the communication data to be transmitted, and sends the encrypted communication data to the receiving end; the receiving end uses the decryption chaotic sequence U'z(t) to decrypt the received encrypted communication data;

[0011] S5, display the decrypted communication data.

[0012] Furthermore, in step S1, the process of using the chaotic system at the transmitting end to generate a multidimensional chaotic sequence includes the following sub-steps:

[0013] S11, select the Lorenz system for encryption verification, the Lorenz equation is:

[0014]

[0015] Where σ = 10, γ = 28, β = 8 / 3, x, y and z represent the x dimension, y dimension and z dimension respectively;

[0016] S12, solve equation (1), set the step length dt = 0.01s, and sample to obtain a number of data;

[0017] S13, select the x dimension as the driving signal and the z dimension as the encrypted signal, scale the sampled data to (0.2, 0.8), set the first 6100 data points as the training set L(t), and the rest of the data as the encrypted data U(t).

[0018] Furthermore, the time delay reserve pool calculation model includes four time delay reserve pools connected in series; each time delay reserve pool includes a masked time delay reserve pool network with 12 virtual nodes and a total output layer.

[0019] Furthermore, in step S2, the process of training the time delay reserve pool calculation model using the training set L(t) includes the following sub-steps:

[0020] S21, substitute the data Lx(t) in the training set L(t) into the time-delay reserve pool calculation model, use equations (2) and (3) to iterate each reserve pool, and obtain the reserve pool state data X of each virtual time node in each time-delay reserve pool:

[0021]

[0022] k = η(x t-τ +γJ) / (1+(x t-τ +γJ) p ) (3);

[0023] Where γ is the input scaling, J is the input after Ux is multiplied by the mask, and there are 12 θ in each τ period, each θ corresponds to a J, x t-τ is the node state at the last time τ, η is the feedback strength; T is the characteristic time; p controls the nonlinearity of the oscillator;

[0024] S22, multiply Lx by mask M to become 12 data after time-division multiplexing input nodes, evolve in the reserve pool, and store the evolution state X i After removing the initial 100 transient points, Lz(t) is taken as the target signal and the output weight matrix W is calculated using the ridge regression formula (4): i :

[0025] W i =L z X i T (X i X i T +λII) -1 (4);

[0026] Where λ is a bias parameter with a size of 1×10-6 to avoid overfitting, and II is the unit matrix;

[0027] S23, combine all the reservoir state data X, and calculate the total output weight matrix W through common ridge regression out , and deployed to the output layer;

[0028] The output of each layer is represented by the W of that layer. i With X i Multiplying them, we get i = 1, 2, 3, 4;

[0029] The final output Uz(t) is given by W out And each layer Xi determines:

[0030] Uz(t)=W out [X1; X2; X3; X4]T (5).

[0031] Furthermore, in step S21, the process of obtaining the reserve pool state data X of each virtual time node in each time delay reserve pool includes the following sub-steps:

[0032] Use Lx(t) to drive the first time-delay reservoir, and adjust its output weight matrix so that the output of the first time-delay reservoir is as close to Lz(t) as possible; use the output of the first time-delay reservoir as the input of the second time-delay reservoir, train the second time-delay reservoir, and make the output of the second time-delay reservoir close to Lz(t); then use the output of the second time-delay reservoir as the input of the third time-delay reservoir, and so on, to complete the training process of the four time-delay reservoirs;

[0033] Get the reserve pool status of each virtual time node in each time-delay reserve pool.

[0034] Further, in step S4, the transmitting end uses the encryption chaotic sequence Uz(t) to encrypt the communication data to be transmitted, and sends the encrypted communication data to the receiving end; the receiving end uses the decryption chaotic sequence U'z(t) to decrypt the received encrypted communication data. The process includes the following sub-steps:

[0035] S41, encrypt the encrypted data h(t) using formula (6):

[0036] S(t)=K1Uz+K2h (6);

[0037] Where Uz is the state variable value of the data U sampled by the chaotic system for encryption at a certain moment, S(t) is the encrypted one-dimensional vector; K1 is greater than K2;

[0038] S42, sending the encrypted signal S(t) to the receiving end;

[0039] S43, using the x dimension in the encryption set U as a driving variable to synchronize the z dimension in the encryption set of the sending end and the receiving end, and decrypting the encrypted signal S(t) using equation (7) at the receiving end:

[0040] H(t)=(S(t)K1-Uz) / K2 (7);

[0041] Among them, K1 and K2 are keys, U(t) is the state variable value of the response system at a certain moment, and H(t) is the one-dimensional vector after decryption.

[0042] Furthermore, K1=1.3, K2=-0.3.

[0043] The present invention proposes a method for realizing secure communication by deploying multi-layer time-delay reserve pool calculation at both ends, that is, configuring a novel multi-layer time-delay reserve pool calculation unit at the transmitting end and the receiving end respectively. After simultaneously training a certain dimensional signal of the chaotic dynamic system for encryption, the reserve pool units at both ends adjust their own output weight parameters, and under the continuous input of a certain training dimensional signal, the stable synchronous output of the encrypted dimensional signal can be achieved, thereby obtaining a useful transmission signal.

[0044] The beneficial effects of the present invention are:

[0045] The invention discloses a confidential communication method based on double-end deployment of multi-layer time delay reserve pool calculation. The double-end deployment of multi-layer time delay reserve pool calculation model can realize high-quality chaotic synchronization between the transmitting end and the receiving end, thereby stably demodulating the useful information transmitted; the system is deployed to ZYNQ7020, and the calculation acceleration is performed by software calling multi-layer time delay reserve pool (TDRs, hardware programmable logic) to obtain a faster chaotic synchronization speed; the on-chip logic is used to realize the driving of ov5640 and HDMI, reduce the use of additional chips, reduce costs and improve reliability. Finally, the Lorenz system is used to realize confidential data transmission through the chaotic masking technology, and the original information can be correctly and effectively restored. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a model structure diagram of an embodiment of the present invention based on ZYNQ7020, wherein APU is Application Processor Unit, which is an ARM Cortex-A9 processor.

[0047] Figure 2 Schematic diagram of a time-delay reserve pool model according to an embodiment of the present invention.

[0048] Figure 3 It is a flow chart of a secure communication method based on dual-end deployment of multi-layer time delay reserve pool calculation according to an embodiment of the present invention.

[0049] Figure 4 This is the effect display in the example of the present invention: Figure 4 In (a), the left screen shows the transmitted image, and the right screen shows the undecrypted received image; Figure 4 In (b), the left screen shows the transmitted image, and the right screen shows the image received after decryption. DETAILED DESCRIPTION

[0050] The present invention will now be described in further detail with reference to the accompanying drawings.

[0051] It should be noted that the terms such as "upper", "lower", "left", "right", "front", "back", etc. cited in the invention are only for the convenience of description and are not used to limit the scope of implementation of the present invention. Changes or adjustments in their relative relationships should be regarded as the scope of implementation of the present invention without substantially changing the technical content.

[0052] Figure 3 1 is a flow chart of a secure communication method based on dual-end deployment of multi-layer time delay reserve pool calculation according to an embodiment of the present invention. Figure 1 and Figure 3 , the confidential communication method comprises the following steps:

[0053] S1, use the chaotic system at the sending end to generate a multidimensional chaotic sequence, generate a training set L(t) containing three dimensions Lx(t), Ly(t) and Lz(t), and send the training set L(t) to the receiving end.

[0054] S2, deploying a time delay reserve pool calculation model consisting of multiple series-connected time delay reserve pools to the receiving end and the sending end at the same time, and using the training set L(t) to train the time delay reserve pool calculation models of the receiving end and the sending end at the same time.

[0055] S3, the chaotic system at the sending end continues to generate new chaotic signals, and the generated new chaotic signals are recorded as the encryption set U(t). U(t) is the continuation of L(x) and contains three dimensions Ux(t), Uy(t) and Uz(t); Ux(t) is retained as the driving set to drive the time-delay reserve pool calculation model at the sending end to obtain the encryption chaotic sequence Uz(t) as the encryption set, and at the same time, Ux(t) is sent to the time-delay reserve pool calculation model at the receiving end to synchronously generate the decryption chaotic sequence U'z(t) consistent with Uz(t) as the solution set.

[0056] S4, the sending end uses the encryption chaotic sequence Uz(t) to encrypt the communication data to be transmitted, and sends the encrypted communication data to the receiving end; the receiving end uses the decryption chaotic sequence U'z(t) to decrypt the received encrypted communication data.

[0057] S5, display the decrypted communication data.

[0058] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0059] This implementation adopts the method of image encryption transmission. Here the system uses the Lorenz system for encryption verification.

[0060] The Lorenz equation is shown in equation (1):

[0061]

[0062] Among them, σ=10, γ=28, β=8 / 3.

[0063] First, solve equation (1), where the step size dt = 0.01s, and sample to obtain sufficiently long data. We select the x dimension as the driving signal and the z dimension as the encrypted signal, scale the data to (0.2, 0.8), and use the first 6100 data points as the training set L(t), and the rest of the data as the encrypted data U(t).

[0064] In order to allow the sender and the receiver to have the same chaotic encryption signal, this invention uses a machine learning method that calculates multiple time delay pools in series to achieve synchronization, that is, the TDRs at both ends are trained using the same training set. First, the time delay pool calculation module is introduced. The basic structure of each time delay pool includes an input layer, an output layer, and a time delay pool network with 12 virtual nodes. The node state is described by equations (2) and (3):

[0065]

[0066] k = η(x t-τ +γJ) / (1+(x t-τ +γJ) p ) (3);

[0067] Where (3) is a delayed-feedback Mackey-Glass oscillator, γ is the input scaling, J is the input after U (L during training) is multiplied by the mask, and there are 12 θ in each τ period, each θ corresponds to a J, X (t-τ) is the node state at the last time τ, and η is the feedback strength. For (2), we use the fourth-order Runge-Kutta formula to obtain a higher-precision X state value.

[0068] First, Lx (1x6100) is multiplied by the mask M (12x1) to become the input node of 12 data after time division multiplexing, and then it is evolved in the reserve pool, and the evolution state X i (12x6100) storage, discarding the initial 100 transient points, using L z The ridge regression calculates the output weight matrix of the first reservoir: W1 (1x12), which is then fed back to the time-delay reservoir. The ridge regression equation is as follows:

[0069] W=L z X T (XX T +λII) -1 (4);

[0070] Where λ is a bias parameter of size 1×10-6 to avoid overfitting, and II is the identity matrix.

[0071] After the first three layers perform ridge regression on Lz one by one, the output weight matrix of each layer is calculated and fed back to the time-delay reserve pool of each layer. The fourth layer does not need an output layer, so ridge regression is not required. After that, the state of the evolved state X (48x6000) after each layer input is used for joint ridge regression calculation Figure 1 The total output weight matrix shown is: W out (1x48) and deployed to the output layer.

[0072] The output of each layer is represented by the W of that layer. i (i=1,2,3,4) and X i (i=1,2,3,4) are multiplied together to obtain.

[0073] The final output U z (t), by W out And each layer Xi (i = 1, 2, 3, 4) determines:

[0074] U z (t) = W out [X1; X2; X3; X4] T ; (5);

[0075] In the encryption and decryption stage: suppose the data to be encrypted h(t) is first used to encrypt the signal to be sent, and the encryption equation is assumed to be as follows:

[0076] S(t)=K1U z +K2h (6);

[0077] Wherein Uz is the state variable value of the data U sampled by the chaotic system at a certain moment for encryption, S(t) is the encrypted one-dimensional vector, and is sent to the receiving end. In order to increase the confidentiality performance, K1 is as large as possible relative to K2. In this embodiment, K1=1.3 and K2=-0.3.

[0078] The x dimension in the encryption set U is used as the driving variable to synchronize the z dimension in the encryption set of the sender and the receiver, and the signal is decrypted to obtain the encrypted information. The decryption equation at the decryption end is:

[0079] H(t)=(S K1-U z ) / K2 (7);

[0080] Among them, K1 and K2 are keys, U(t) is the state variable value of the response system at a certain time, and H(t) is the one-dimensional vector after decryption. The test results are as follows: Figure 4 shown.

[0081] The design is implemented on the Zynq-7020 chip, which integrates a dual-core ARM Cortex-A9 processor and an FPGA programmable logic unit into a single chip, thus forming a PS (Processing System) plus PL (Programmable Logic) single-chip SoC solution. The chip's PS system integrates two ARM Cortex-A9 processors, AMBA interconnect, internal memory, external memory interface and peripherals. The ZYNQ7020 transmission system is mainly composed of 0v5640 camera acquisition, HDMI external display display, TDR module, PL multiplication acceleration, ZYNQ 7020 Processing System, and serial port. See the operation process for details. Figure 3 Flowchart sending end. The ZYNQ7020 receiving system is mainly composed of HDMI display, TDR module, PL multiplication acceleration, and ZYNQ 7020 Processing System. See the operation flow Figure 3 Flowchart receiving end. Both the sending and receiving ends are trained and synchronized through TDR (IP). TDR is written in HLS C++ and calls the HLS acceleration interface to fully utilize resources to complete the design.

[0082] This embodiment mentions a method for implementing a confidential communication FPGA with dual-end deployment of multi-layer time-delay reservoir (TDR) calculations, which simultaneously deploys multiple series time-delay reservoir calculations to the receiving end and the transmitting end, and implements a new type of chaotic synchronization mechanism based on FPGA design, which can be used for confidential communication of various types of information. This method can achieve high-quality chaotic synchronization between the transmitting end and the receiving end, thereby stably demodulating the transmitted useful information. In terms of specific implementation, this communication mechanism can utilize the ZYNQ7020 SOC development board, and accelerate the calculation by calling the multi-layer time-delay reservoir calculation through software to obtain a faster chaotic synchronization speed; make full use of the on-chip logic to realize the drive of ov5640 and HDMI, reduce the use of additional chips, reduce costs and improve reliability. Finally, the mechanism uses the Lorenz system for chaotic encryption, realizes the confidential transmission of data, and correctly and effectively restores the original information.

[0083] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A secure communication method based on double-end deployment of multi-layer time delay reserve pool calculation, characterized in that: The secure communication method comprises the following steps: S1, using the chaotic system at the sending end to generate a multidimensional chaotic sequence, generating a training set L(t) containing three dimensions Lx(t), Ly(t) and Lz(t), and sending the training set L(t) to the receiving end; S2, deploying a time delay reserve pool calculation model consisting of multiple series-connected time delay reserve pools to the receiving end and the sending end at the same time, and using the training set L(t) to train the time delay reserve pool calculation models of the receiving end and the sending end at the same time; S3, the chaotic system at the sending end continues to generate new chaotic signals, and the generated new chaotic signals are recorded as the encryption set U(t). U(t) is the continuation of L(x) and contains three dimensions Ux(t), Uy(t) and Uz(t); Ux(t) is retained as the driving set to drive the time-delay reserve pool calculation model at the sending end to obtain the encryption chaotic sequence Uz(t) as the encryption set, and at the same time, Ux(t) is sent to the time-delay reserve pool calculation model at the receiving end to synchronously generate the decryption chaotic sequence U'z(t) consistent with Uz(t) as the solution set; S4, the sending end uses the encryption chaotic sequence Uz(t) to encrypt the communication data to be transmitted, and sends the encrypted communication data to the receiving end; the receiving end uses the decryption chaotic sequence U'z(t) to decrypt the received encrypted communication data; S5, display the decrypted communication data.

2. The secure communication method based on double-end deployment of multi-layer time delay reserve pool calculation according to claim 1 is characterized in that: In step S1, the process of using the chaotic system at the transmitting end to generate a multidimensional chaotic sequence includes the following sub-steps: S11, select the Lorenz system for encryption verification, the Lorenz equation is: Where σ = 10, γ = 28, β = 8 / 3, x, y and z represent the x dimension, y dimension and z dimension respectively; S12, solve equation (1), set the step length dt = 0.01s, and sample to obtain a number of data; S13, select the x dimension as the driving signal and the z dimension as the encrypted signal, scale the sampled data to (0.2, 0.8), set the first 6100 data points as the training set L(t), and the rest of the data as the encrypted data U(t).

3. The secure communication method based on double-end deployment of multi-layer time delay reserve pool calculation according to claim 1 is characterized in that: The time delay reserve pool calculation model includes four time delay reserve pools connected in series in sequence; each time delay reserve pool includes a masked time delay reserve pool network with 12 virtual nodes and a total output layer.

4. The secure communication method based on double-end deployment of multi-layer time delay reserve pool calculation according to claim 3 is characterized in that: In step S2, the process of training the time delay reserve pool calculation model using the training set L(t) includes the following sub-steps: S21, substitute the data Lx(t) in the training set L(t) into the time-delay reserve pool calculation model, use equations (2) and (3) to iterate each reserve pool, and obtain the reserve pool state data X of each virtual time node in each time-delay reserve pool: k=η(x t-τ +γJ) / (1+(x t-τ +γJ) p ) (3); Where γ is the input scaling, J is the input after Ux is multiplied by the mask, and there are 12 θ in each τ period, each θ corresponds to a J, x t-τ is the node state at the last time τ, η is the feedback strength; T is the characteristic time; p controls the nonlinearity of the oscillator; S22, multiply Lx by mask M to become 12 data after time-division multiplexing input nodes, evolve in the reserve pool, and store the evolution state X i After removing the initial 100 transient points, Lz(t) is taken as the target signal and the output weight matrix W is calculated using the ridge regression formula (4): i : W i =L z X i T (X i X i T +λII) -1 (4); Where λ is a bias parameter with a size of 1×10-6 to avoid overfitting, and II is the unit matrix; S23, combine all the reservoir state data X, and calculate the total output weight matrix W through common ridge regression out , and deployed to the output layer; The output of each layer is obtained by multiplying Wi and Xi of the layer, i = 1, 2, 3, 4; The final output Uz(t) is given by W out And each layer Xi determines: Uz(t)=W out [X1;X2;X3;X4] T (5)。 5. The secure communication method based on double-end deployment of multi-layer time delay reserve pool calculation according to claim 4 is characterized in that: In step S21, the process of obtaining the reserve pool state data X of each virtual time node in each time delay reserve pool includes the following sub-steps: Use Lx(t) to drive the first time-delay reservoir, and adjust its output weight matrix so that the output of the first time-delay reservoir is as close to Lz(t) as possible; use the output of the first time-delay reservoir as the input of the second time-delay reservoir, train the second time-delay reservoir, and make the output of the second time-delay reservoir close to Lz(t); then use the output of the second time-delay reservoir as the input of the third time-delay reservoir, and so on, to complete the training process of the four time-delay reservoirs; Get the reserve pool status of each virtual time node in each time-delay reserve pool.

6. The secure communication method based on double-end deployment of multi-layer time delay reserve pool calculation according to claim 1 is characterized in that: In step S4, the transmitting end uses the encryption chaotic sequence Uz(t) to encrypt the communication data to be transmitted, and sends the encrypted communication data to the receiving end; the receiving end uses the decryption chaotic sequence U'z(t) to decrypt the received encrypted communication data. The process includes the following sub-steps: S41, encrypt the encrypted data h(t) using formula (6): S(t)=K1Uz+K2h (6); Where Uz is the state variable value of the data U sampled by the chaotic system for encryption at a certain moment, S(t) is the encrypted one-dimensional vector; K1 is greater than K2; S42, sending the encrypted signal S(t) to the receiving end; S43, using the x dimension in the encryption set U as a driving variable to synchronize the z dimension in the encryption set of the sending end and the receiving end, and decrypting the encrypted signal S(t) using equation (7) at the receiving end: H(t)=(S(t)K1-Uz) / K2 (7); Among them, K1 and K2 are keys, U(t) is the state variable value of the response system at a certain moment, and H(t) is the one-dimensional vector after decryption.

7. The secure communication method based on double-end deployment of multi-layer time delay reserve pool calculation according to claim 6 is characterized in that: K1=1.3, K2=-0.3.

Images