Elman-Based Secure Data Transmission Quality Prediction for Complex IoT Networks

. The information age has brought earth-shaking changes. For interconnection of all things, the data transmission has widely employed the Internet of Things (IoT). The IoT transmission faces complex environments. The secure data transmission is very important for mobile IoTnetworks. The secure data transmission quality prediction is investigated for mobile IoTnetworks. The probability of strictly positive secrecy capacity (SPSC) is used to evaluate the secure data transmission quality, and the expressions are ﬁrst derived. Then, employing Elman network, a secure data transmission quality intelligent prediction approach is proposed. The extensive simulations are run to evaluate the proposed approach. The simulation results show that the Elman-based approach can achieve a higher quality precision than other methods. The Elman-based approach also can achieve a lower time complexity.


Introduction
With the explosive growth of mobile applications, Internet of things (IoT) networks are widely used to transmit data [1].
e fifth generation (5G) mobile communication also has been widely used in mobile IoT networks [2,3].Different 5G applications widely use sea-land-air mobile communication networks [4,5].e global and diversified application will provide quick and convenient services for IoT users.However, due to IoT mobility and the diversity of IoT networks, the physical layer security (PLS) of 5G mobile IoT networks is facing many challenges [6].
PLS of 5G IoT networks is a research hot spot [7].Lowcomplexity schemes for IoT PLS were presented in [8].In [9], power control mechanism and antenna transmission scheme were used to realize the secure data transmission in cognitive wiretap networks.Considering the mobile healthcare networks, Xu et al. [10] investigated the PLS performance using the deep learning method.In [11], the authors used compressed sensing and cooperative schemes to achieve the secure transmission.Considering the user and relay selection, Fan et al. [12] analyzed two criteria and investigated the achievable PLS performance.e authors of [13] analyzed the upper and lower bounds on PLS performance over dependent fading channels.
e IoT data transmission faces a wide variety of scenarios and complex environments.
e PLS issue is more and more serious.However, predicting and evaluating the secure data transmission quality are very difficult.Recently, machine learning techniques are applied in 5G wireless communications [14,15].In medical IoT, support vector machine (SVM) model was used to train data privacy [16].High-performance visual tracking was achieved by an extreme learning machine (ELM) model in [17].In [18], general regression (GR) model was used to evaluate the video transmission quality.e radial basis function (RBF) network was optimized to reconstruct the image in [19].
e studies of secure data transmission quality prediction are rare.So, our paper investigates the secure data transmission quality prediction of mobile IoT networks.e main contributions are given as follows.
(1) With amplify-and-forward (AF) relaying scheme, we use SPSC to evaluate secure data transmission quality and derive the exact expressions.(2) To realize real-time analysis of secure data transmission quality, we propose a secure data transmission quality prediction approach based on the Elman neural network.e proposed approach is compared with ELM, GR, and RBF methods.
(3) rough the extensive simulations, we verify the derived results.Compared with different methods, the quality assessment effect of Elmanbased approach is better, and time complexity is lower.

The IoT System Model
e system has a mobile source (S), mobile destination (D), mobile eavesdropper (E), and mobile relay (R). Figure 1 shows the system model.
First, MR receives the signal r SR as where w SR is Gaussian noise.
In the second time slot, D and E receive the signals r Rk , k ∈ {D, E}, as e received SNR W SRk is given as where W SRk is very complex.We approximate W SRk as [22] Bloch et al. [23] give the instantaneous secrecy capacity as

Secure Data Transmission Quality Analysis
e probability of SPSC F SPSC is used to evaluate the secure data transmission quality.We will give the analysis.
According to the (6), F SPSC is given as With the help of [24], we obtain the PDF and CDF of W SRAk as follows: Substituting ( 8) and ( 9) into ( 7), F SPSC is expressed as (11)

Secure Data Transmission Quality Prediction Approach
4.1.Data Sets.T i � (X i , y i ). e input X i includes 5 indicators.X i is given as e output y i is the SPSC.By using (11), the corresponding y i can be obtained.

Predictive Evaluation.
For PP testing data, MSE and AE are used to evaluate the prediction effect:

Numerical Results
In this section, E � 1 and μ � W RD /W RE (in decibels).With parameters in Table 1, we evaluate the SPSC performance with c � 10 dB in Figure 3. Simulation results show the following: (1) increasing u improves the SPSC 2 Mobile Information Systems performance; (2) for Nakagami channels, the secure data transmission quality is the best.is is because a higher u improves the S ⟶ R⟶ D channel while degrading the S⟶R⟶E channel.
In Figures 4-11, ELM, GR, and RBF methods are compared with the Elman approach.Table 2 gives the simulation parameters.e MSE and AE of Elman approach are 0.00014 and 0.011, which are the lowest MSE and AE in the five methods.is is because Elman is a typical dynamic recurrent neural network and can adapt to the time-varying characteristics by adding a context layer.
e MSE is compared in Figure 12.Compared with GR, Elman has a better MSE performance, but the running time is longer than GR.Furthermore, compared with other methods, Elman has a higher quality precision and a lower time complexity.
Output Layer The 1st time slot The 2nd time slot

Conclusion
is paper investigated the SPSC prediction of mobile IoT Networks.
e exact expressions for SPSC were derived.Furthermore, based on the Elman network, we proposed an intelligent secure data transmission quality prediction algorithm.e theoretical analysis showed the following: (1) the SPSC performance over Nakagami channels was the best; (2) compared with different methods, the Elman algorithm can achieve a higher quality precision.

Figure 1 :Figure 3 :Figure 4 :Figure 5 :Figure 6 :Figure 7 :Figure 8 :Figure 9 :Figure 10 :Figure 11 :
Figure 1: System model.E is the transmission power.For R and S, E is allocated with K. e channel coefficient h is 2-Nakagami distribution [20, 21].M SR , M RD , and M RE are the relative geometrical gains of S ⟶ R, R ⟶ D, and R ⟶ E links, respectively.

Table 2 :
e parameters for different methods.