Research on security encryption mechanism of physical Layer based on iterative quantization method

Abstract

Third-party eavesdropping is a unsolved problem in the process of data transmission in the physical layer of IoT (Internet of Things) in Power Systems. The security encryption effect is affected by channel noise and the half-duplex nature of the wireless channel, which leads to low key consistency and key generation rate. To address this problem, a reliable solution for physical layer communication security is proposed in this paper. First, the solution improved the key consistency by dynamically adjusting the length of the training sequence during feature extraction; Second, using an iterative quantization method to quantify the RSS (Received Signal Strength) measurements to improve generation rate of the key. Finally, based on the short-time energy method for the extraction of wireless frame interval features, by monitoring the change of inter-frame interval features, we can quickly determine whether there is an eavesdropping device into the link. Simulation results show that the reciprocity of legitimate channels R (R will be explained in detail in the following) is improved by 0.1, the key generation rate is increased by about 70%, and the beacon frames are extracted from the wireless link with good results compared to the methods that do not use dynamic adjustment of the pilot signal during the channel probing phase. The result shows that this method can effectively prevents third-party eavesdropping, effectively improves the key consistency and generation rate, and effectively implements beacon frame detection.

Keywords

Security encryption mechanism iterative quantization physical layer keys wireless frame interval

1. Introduction

In recent years, with the continuous development of big data, cloud computing, Internet of Things and other mobile Internet technologies, the new digital technology and traditional power technology deep integration of the new power system has become an effective means to promote the green transformation of energy and power. Wireless communication technology has gradually replaced the traditional wired network with cumbersome wiring and high cost due to its strong flexibility and expandability, and has been widely used in the field of power IoT [1]. The open nature of the wireless communication link leads to its communication process is very vulnerable to third-party eavesdropping, communication data loss or tampering and other data security threats, so the new power system communication security can not be separated from the support and protection of cryptographic technology and individual device identification technology. Accurate identification of illegal accessors in wireless networks is of great significance for maintaining communication security. The current research work on traditional individual device identification technology mainly focuses on the extraction of transient and steady-state characteristics of wireless signals, which has the drawbacks of low precision and low practicality [2]. For cryptography, the degree of security of traditional network layer encryption algorithms generally depends on the difficulty of solving certain mathematical problems, and with the development of technology, the security of communication cannot be sufficiently guaranteed [3]. At the same time, for open wireless communication networks, if traditional key algorithms are used, there is a high risk of session keys being intercepted during key distribution, and communication data can be easily stolen. Moreover, traditional encryption algorithms generally require huge computational effort, which is difficult to implement for wireless sensor networks with limited computational power and small storage space.

To address these issues, a wireless channel-based physical layer key generation technique has been proposed, which has lower complexity and does not rely on a trusted third party for key distribution compared to traditional encryption algorithms such as RSA [4]. By physical layer security, we mean using the uniqueness and reciprocity of physical channels to encrypt information, generate passwords, identify legitimate users, etc. The two most critical aspects of the physical layer key generation are channel detection and channel feature utilization [5]. Channel detection is the process of extracting the wireless channel features, and channel feature utilization is the process of using the extracted channel features as the initial key source and obtaining the session key after a series of protocols and transformations. Commonly used channel characteristics include Received Signal Strength (RSS), Channel State Information (CSI), Channel Impulse Response (CIR), etc. [6]. Among them, RSS is widely used in the study of key extraction because it does not require too much equipment and can be used directly in wireless devices.

In fact, most of the commonly used wireless communication channels are half-duplex systems, and coupled with environmental factors such as Gaussian white noise, the channel characteristics extracted by the two communicating parties are hardly identical, which leads to the problem of inconsistent keys. In order to enhance channel reciprocity and improve key consistency, many scholars have tried to use filters to reduce noise. literature [7] proposed a light weight physical-layer security-based authentication and key generation scheme The Internet-of-Things (IoT) will connect different types of cyber physical systems that are supported by different wireless transmission technologies. literature [8] proposed a novel method called MLRefine to counter these effects by refining RSS measurement data to obtain more accurate values that can enhance ranging and localization accuracies. Cubic Farrow filters for interpolation were implemented in the literature [9], respectively, to solve the non-simultaneous measurement problem. Although all these methods improve key consistency to some extent, most of the quantization methods they use are rounds of lossy quantization, and wireless devices discard a large number of channel feature measurements during key quantization, resulting in a low key generation rate. In order to improve the key generation rate, Xiao et al. [10] proposed a multi-relay-assisted key generation algorithm with optimal relay selection algorithm using optimal power allocation algorithm, which improves the key generation rate, but the method is extremely demanding on the environment. Chen et al. [11] proposed an algebraic grouping based key generation scheme using polynomial time compiled code method based on polar coordinate codes, which improves the key generation rate, but the key consistency is low and the operation is difficult. Physical non cloning function (PUF) is a hardware security technology, which uses inherent device changes to generate non cloning unique device response to a given input. However, PUF depends on small manufacturing differences, and measurement accuracy is a major challenge. Manufacturing differences will lead to inconsistent equipment. The literature [12] uses a power allocation scheme to increase the rate of secret key generation while making it more difficult for an eavesdropper to extract the key between the communicating parties, but it lacks in consistency.

At the same time, in order to make up for the shortcomings of low accuracy and poor practicality of traditional individual device identification technology, we have researched individual wireless network device identification technology based on wireless frame interval and proposed a beacon inter-frame interval extraction technology based on the short-time energy method, which lays the foundation for further identifying third-party devices that invade wireless networks by monitoring the wireless link frame interval characteristics and avoiding the risk of data being stolen or tampered with during the transmission process from the root [13].

In order to simultaneously improve the key consistency and key generation rate, and to further prevent the risk of data loss or tampering during transmission based on the encryption protection of data, this paper combines the link inter-frame interval feature extraction technique with the physical layer key generation technique to propose a reliable method for secure data transmission protection at the side end. The main contributions of this paper are as follows.

1.
The channel detection time is controlled within the channel coherence time as much as possible by dynamically adjusting the length of the guide frequency training sequence during the channel detection phase, thus ensuring that the channel features extracted by both sides of the communication are highly consistent.
2.
propose an iterative quantization algorithm to quantize RSS measurements in multiple rounds during the quantization phase in order to improve the key generation rate and the utilization of RSS measurements.
3.
propose a wireless link inter-frame interval extraction method based on the short-time energy method to monitor the change of the wireless link frame interval characteristics by extracting the set of beacon inter-frame intervals, and determine whether a third-party device has invaded the link by the change of the link frame interval characteristics.

Figure 1.
Eavesdropping channel model.

Figure 2.
Physical layer key generation flow chart.

2. Theoretical foundation

2.1 Channel model

The eavesdropping channel model was born in 1975 [14], it can be represented by Fig. 1. In the figure, Alice and Bob are the legitimate communication parties who transmit information through the public channel, and Eve is the third party who eavesdrops information through the eavesdropping channel.

According to the short-time reciprocity principle of the common channel, two legitimate communication nodes Alice and Bob can obtain the same channel impulse response and extract the channel features with great similarity from them if they communicate at the same frequency at the same time, so that the extracted channel features can be used to generate relatively consistent session keys [15]. The process of physical layer key generation consists of five main stages: channel probing, feature extraction, quantization, message negotiation and confidentiality enhancement. The physical layer key generation process is shown in Fig. 2.

2.2 Channel probing and feature extraction

Channel probing and feature extraction is the first step of physical layer key generation, in which both legitimate communicating parties detect the wireless channel by sending guide signals to each other to extract channel characteristics. During the channel coherence time, Alice and Bob obtain the same common channel response, while if the distance between the eavesdropping user Eve and the legitimate communication nodes Alice and Bob is more than half of the electromagnetic wavelength, he cannot obtain the channel characteristics similar to the legitimate nodes.

Alice and Bob perform channel probing by sending a training sequence to each other. Using $s(t)$ to represent this training sequence, the signal received by Bob from Alice can be represented by Eq. (1).

$\displaystyle y_{AB}=m_{AB}\times s(t)+n_{AB}(t)$ (1)

The signal received by Alice from Bob can be represented by Eq. (2).

$\displaystyle y_{BA}=m_{BA}\times s(t)+n_{BA}(t)$ (2)

Where $m_{AB}$ and $m_{BA}$ are the channel response from Alice to Bob and Bob to Alice at that moment, respectively, and $n_{AB}(t)$ and $n_{BA}(t)$ denote the Gaussian white noise of the wireless channel at that moment.

The signals from Alice and Bob eavesdropped by the third party Eve can be represented by Eqs (3) and (4), respectively.

$\displaystyle y_{AE}=m_{AE}\times s(t)+n_{AE}(t)$ (3) $\displaystyle y_{BE}=m_{BE}\times s(t)+n_{BE}(t)$ (4)

Similarly, $m_{AE}$ and $m_{BE}$ denote the channel responses from Alice to Eve and Bob to Eve at that moment, respectively. $n_{AE}(t)$ and $n_{BE}(t)$ denote the Gaussian white noise of the two eavesdropping channels at that moment, respectively.

It can be seen that the consistency of the signals received by both Alice and Bob depends mainly on $m_{AB}$ and $m_{BA}$ , however, due to the half-duplex nature of most current wireless channels, Alice and Bob cannot detect them simultaneously, and coupled with the fact that channel reciprocity is affected by noise and dynamically changing environments, differences in $m_{AB}$ and $m_{BA}$ are inevitable. Therefore, both communicating parties must keep the time required to complete the two-way channel probing within the channel coherence time in order to ensure that the channel features extracted by both parties are highly consistent. The channel coherence time is inversely proportional to the maximum Doppler shift $f_{\max}$ , which can be expressed by Eq. (5) [16].

$\displaystyle T_{C}=\frac{1}{f_{\max}}$ (5)

When communication nodes send training sequences to each other for channel probing, if the time required to send the sequences is greater than the coherence time, it will lead to a situation where the signal is completely changed. It can be seen that the length of the channel coherence time determines the value of the time interval of the training sequence sent by the communication nodes of the channel probing process, which affects the length of the training sequence. Therefore, if we want to ensure the reciprocity of wireless channels at the root, one of the methods is to dynamically adjust the length of the training sequence as the channel coherence time changes, which can also reduce the workload of subsequent key consistency correction.

Use $t_{AB}$ to denote the time required for two-way probing by the communicating parties and $T_{C}$ to denote the coherence time, and $\alpha=\frac{t_{AB}}{T_{C}}$ denotes the ratio of the two, which is related to the channel reciprocity. When $\alpha>1$ , it means that the channels are not correlated; when $0\leqslant\alpha\leqslant 1$ , it means that the wireless channels are reciprocal, and the smaller $\alpha$ is, the higher the channel reciprocity is. The time $t$ required to send the training sequence between the communicating parties should also be less than $t_{AB}$ .

Let the initial sequence length be $L$ , the sending rate be $S_{t}$ , and the initial sending time be $t_{0}=L/S_{t}$ . The process of dynamically adjusting the length of the training sequence is as follows.

Alice sends the initial training sequence to Bob, who receives it and calculates the initial coherence time $T_{C}$ using Eq. (5).

Bob computes the current $t_{AB}$ by substituting the $T_{C}$ computed in the first step into the equation $t_{AB}=\alpha\ast T_{C}$ .

If $2t_{0}<t_{AB}$ , then no adjustment of the initial sequence is required; otherwise the initial training sequence length is reduced to ${L}^{\prime}$ such that $t<t_{AB}$ .

Bob requests Alice to resend the training sequence of length ${L}^{\prime}$ and performs channel estimation accordingly.

Alice and Bob repeat the above process until the adjusted training sequence length can complete the channel estimation in coherent time. In this paper, we use the least squares method for channel estimation.

2.3 Iterative quantification

Received signal strength (RSS) is the easiest channel feature to extract, and is obtained by sending a fixed signal at the transmitter and perform modulo operation on it at the receiver after receiving it. The quantization phase is the process of converting the RSS values which is extracted by Alice and Bob into a bit stream. Keeping Alice and Bob’s two-way probes in channel coherence time, assume that the sequence of RSS measurements collected by Alice is $X[{x_{1},x_{2},x_{3},\ldots,x_{n}}]$ and the sequence of RSS measurements collected by Bob is $Y[{y_{1},y_{2},y_{3},\ldots,y_{n}}]$ . The mean and variance of $X$ and $Y$ are taken and denoted as $\mu_{k}$ and $\sigma_{k}$ , respectively, where k denotes the number of rounds of iterative quantization. Alice and Bob used Eq. (6) to calculate the quantization thresholds $q_{k}^{+}$ and $q_{k}^{-}$ .

$\displaystyle\left\{{{\begin{array}[]{l}{q_{k}^{+}=\mu_{k}+\delta\times\sigma_% {k}}\\ {q_{k}^{-}=\mu_{k}-\delta\times\sigma_{k}}\\ \end{array}}}\right.$ (6)

where $\delta\in({0,1})$ denotes the quantization factor, and its value is related to the channel reciprocity and signal-to-noise ratio. When the channel reciprocity is low, the value of $\backslash$ delta increases, the interval range of $q_{k}^{+}$ and $q_{k}^{-}$ increases, and the probability of the RSS measurement being quantified decreases; on the contrary, when the channel reciprocity is strong, the probability of the RSS measurement being quantified is enhanced, and the key generation rate increases.

Define the quantizer $R(x)$ , expressed by Eq. (7).

$\displaystyle R(x)=\left\{{{\begin{array}[]{ll}0&x\leqslant q^{-}\\ 1&x\geqslant q^{+}\\ \end{array}}}\right.$ (7)

Alice and Bob use the RSS values as input to the quantizer $R(x)$ , whose output is the quantized bitstream. When the RSS value is greater than $q_{k}^{+}$ , it is quantified as 1; when the RSS value is less than $q_{k}^{-}$ , it is quantified as 0. The measurements between $q_{k}^{+}$ and $q_{k}^{-}$ , which are susceptible to noise, are used as input values for the next round of iterative quantization, and quantized cyclically until all RSS measurements are converted to bit streams. And since Eve node cannot get the same RSS measurement as the legitimate node, it cannot get the session key consistent with the legitimate node by quantifying the RSS value.

2.4 Key generation

In real application scenarios, the noise of the wireless channel is unavoidable, and the initial session key quantified by both communicating parties is not exactly the same due to Gaussian white noise and hardware facilities, so Alice and Bob need to negotiate its consistency. This paper uses the Cascade code combined with error correction code technique for negotiation [17].

First, Alice and Bob divide the initial bit stream obtained by quantization into multiple sequences of equal length, and perform parity check on each sequence. Second, the two communicating parties compare the parity bits of each sequence, and if a pair of sequences does not match, Alice and Bob respectively divide the sequence into two equal-length bitstream groups, then parity check them, send the obtained parity bits to each other through the common channel, and choose the next operation according to the received parity bit comparison result. If the sequence checksum is different, delete the last bit of the sequence and repeat the above operation for the remaining bit stream until the checksum of the subsequence is the same.

At this point it is not secure to use the bit stream obtained after consistent negotiation between the communicating parties as the session key. During the consistency negotiation phase, the communicating parties Alice and Bob use the common channel several times for information transmission, and Eve, the eavesdropper, can then infer the bit stream obtained by Alice and Bob quantizing the RSS measurements based on the information eavesdropped from the eavesdropping channel. Therefore, the bit stream obtained after the consistency negotiation also needs to be confidentiality-enhanced to get to the final session key. In this paper, we use a Double hash function to enhance the secrecy of the negotiated bit stream to obtain the session key [18], which is not described here.

2.5 Wireless link beacon frame spacing extraction

The prerequisite for obtaining the wireless frame spacing characteristics is to obtain the location of the frame starting point. The commonly used methods for detecting the starting point of a frame are Bayesian Online Changepoint Detection method [19], Permutation Entropy method [20] and short-time energy method [21].

The advantage of the short-time energy method is that it effectively improves the signal-to-noise ratio and can extract the onset of vibration events from the complex noise environment of the site.We use the short-time energy method to detect the starting point of the frame. In the IEEE802.11 standard, wireless routers (APs) broadcast beacon frames with a period of 100 ms. In practice, wireless APs usually interact with multiple stations for information, so it is necessary to reject non-beacon frames when a single wireless AP is in operation, so as to calculate the frame interval of that wireless AP in actual transmission, and then to study its characteristics.

The wireless AP link beacon frame interval extraction process is as follows.

1.
Assuming that the signal is $x(t)$ , which is normalized to the interval [ $-$ 1, 1] according to Eq. (8).

$\displaystyle{x(t)}=2\times\frac{x(t)-\min({x(t)})}{\max({x(t)})-\min({x(t)})}-1$ (8)
2.
Choose a suitable window width size and use the short-time energy method to find the energy curve of the signal, whose short-time energy can be expressed by Eq. (9), where $g(n)$ denotes the rectangular window function.

$\displaystyle E_{n}=\mathop{\sum}\limits_{k=n-N+1}^{\infty}[{x(k)g({n-k})}]^{2}$ (9)
3.
Set the threshold T to find the value of the energy mutation point. The band with energy value greater than or equal to T can be regarded as having wireless frame transmission and its amplitude is set to 1. The band with energy value less than T can be regarded as a blank band and its amplitude is set to 0.
4.
Count the number of samples corresponding to the segment with wireless frame transmission, and calculate the duration of this wireless frame based on the sampling rate. Assuming that the $i$ -th frame has $n_{i}$ sampling points, the duration $t_{i}$ of the wireless frame can be calculated from Eq. (10), where $f_{s}$ represents the sampling rate.

$\displaystyle t_{i}=\frac{n_{i}}{f_{s}}$ (10)
5.
The beacon frames are filtered based on the frame duration calculated in step 4. In IEEE 802.11 standard, the longest frame duration of beacon frames can be maintained in the range of 1 $\sim$ 3 ms. In this paper, we use 1 ms and 3 ms as the boundary to separate the non-beacon frames to complete the screening of wireless frames.
6.
Calculate the set of frame interval of adjacent beacon frames based on the starting position of each beacon frame.

According to the set of wireless frame intervals extracted by the above steps can further identify individual wireless APs, quickly monitor third-party eavesdropping devices in wireless communication, and further ensure the security of data transmission at the side end on the basis of data encryption. In this paper, only the wireless frame interval extraction is studied, and the subsequent steps are not described.
3. Simulation experiments and performance analysis

This paper uses MATLAB for simulation experiments. Both the public channel model and the eavesdropping channel model in the experiment use the SISO Gaussian channel model with a signal-to-noise ratio of 20 db, where the Gaussian white noise present has a mean value of 0 and a variance of $\sigma_{n}^{2}$ . Suppose there exist three communication devices Alice, Bob and Eve, where Alice and Bob are the nodes on both sides of the legitimate communication and Eve is the eavesdropping node. Assume that the length of the initial guide frequency training sequence sent by Alice and Bob for channel probing is 1000. R is used to represent the correlation of the channels used by the communicating parties, and the mutual covariance function is introduced, whose expression is shown in Eq. (11) [22].

$\displaystyle R_{AB}=\frac{C_{y_{AB}y_{BA}}(\tau)}{\sqrt{C_{y_{AB}y_{AB}}(0)C_% {y_{BA}y_{BA}}(0)}}$ (11)

Where $y_{AB}$ and $y_{BA}$ denote the channel probing results for Alice and Bob respectively, and $R_{AB}$ denotes the reciprocity of the common channel. Similarly, the correlation of eavesdropping channels between Eve and Alice and Bob can be represented by $R_{AE}$ and $R_{BE}$ , respectively.

From Eq. (8), it can be seen that the closer $R_{AB}$ is to 1, the more similar the channel probing results between the two nodes are, indicating that the higher the correlation, the higher the consistency of the generated keys; the closer $R_{AB}$ it is to 0, the lower the correlation, the lower the consistency of the generated keys. The closer $R_{AE}$ and $R_{BE}$ are to $R_{AB}$ , the more similar the channel probing result of the eavesdropping node Eve is to that of the legitimate node, the more it can generate a session key similar to that of the legitimate node, and the lower the communication security; conversely, the greater the difference between $R_{AE}$ and $R_{BE}$ and $R_{AB}$ , the higher the communication security.

This experiment uses a wireless channel with a signal-to-noise ratio of 20 db, assuming that the $\alpha$ value of the channel is 0.5, and a total of 200 simulations are collected to compare the data. In each experiment, the maximum Doppler shift $f_{\max}$ of size 10 to 50 Hz is randomly selected and $T_{C}$ changes accordingly. The simulation results of the coherence degree $R_{AB}$ of the legitimate communication nodes Alice and Bob and the coherence degree $R_{AE}$ and $R_{BE}$ of the eavesdropping nodes Eve and Alice and Bob are shown in Fig. 3.

Figure 3.

Public channel versus eavesdropping channel reciprocity diagram.

As can be seen from Fig. 3, the value of $R_{AB}$ fluctuates around 0.95, indicating that the correlation of the two communication nodes is high and the channel probing results tend to be consistent; while the values of $R_{AE}$ and $R_{BE}$ fluctuate more and do not exceed 0.4, which are significantly different from $R_{AB}$ , indicating that the coherence between Eve and Alice and Bob is poor and the channel The detection results are also very different, which verifies the security feasibility of the method in this paper.

Keeping the above simulation experiment environment unchanged, the detection method using fixed training sequence is compared with the detection method of dynamically adjusting the training sequence proposed in this paper, each time the length of the current training sequence is dynamically adjusted to 0.8 times of the last firing sequence, and a total of 200 simulation experiments are conducted, and the experimental results are shown in Fig. 4.

As can be seen from Fig. 4, the channel reciprocity $R_{AB}$ fluctuates considerably when the traditional detection method with fixed sequence length is used, with the minimum value even tending to 0.8, while $R_{AB}$ tends to be smooth and fluctuates around 0.95 when use the method which is proposed in this paper. The results show that the channel probing method proposed in this paper effectively improves the channel reciprocity and the performance is relatively stable, which can improve the key consistency of both sides of the communication from the root.

During the channel probing process, Alice and Bob lose about 1 packet for every 100 packets sent, i.e., the packet loss rate is about 1%. The iterative quantization scheme proposed in this paper can quantize each RSS measurement to 1 bit. Using $T_{R}$ to denote the total number of packets sent by both sides, $D_{R}$ to denote the total number of mismatched bits, and $L_{R}$ to denote the total number of lost packets, the key generation rate KGR can be expressed by Eq. (12).

$\displaystyle\textit{KGR}=\frac{T_{R}-D_{R}-L_{R}}{T_{R}}$ (12)

And the Key Disagreement Rate (KDR) is the ratio of the number of mismatched bits in the generated keys of the communicating parties to the total number of bits in the generated keys. In this scheme, the main factor affecting the KDR is the parameter $\delta$ of the iterative quantization algorithm. Take the quantization factor $\delta$ value 0.01, when the key mismatch rate is the lowest. Different values of $T_{R}$ are set and multiple simulation experiments are performed using the iterative quantization method proposed in this paper to calculate the KGR and KDR for each experiment. Table 1 shows the experimental results.

Table 1

Table of experimental results of KGR and KDR using iterative quantification method

$T_{R}$	100	150	220	300	360	440	500
$D_{R}$	6	8	11	15	18	21	23
KGR	93%	93.5%	94%	94%	94%	94.2%	94.4%
KDR	4.4%	4.2%	4%	4.2%	4.1%	4.1%	4.2%

Figure 4.

Experimental results comparing the fixed sequence length detection method with the method in this paper.

As can be seen from Table 1, the KGR value always remains above 93% and increases with the increase of communication packets, while the KGR of the traditional round of lossy quantization method is only 20%–30%. And KDR are also kept in the low range. The experiments verify that the iterative quantization method proposed in this paper significantly outperforms the traditional one-round lossy quantization method and greatly improves the utilization of bits, thus increasing the key generation rate.

Use the randomness test tool provided by NIST (the National Institute of Standards and Technology) to test the randomness of the keys. The result of each test item is the corresponding p-value, which takes the range of (0, 1), when $p$ is closer to 1, it means that the key is more random, and when $p>0.01$ , it means that the key has better randomness. Table 2 shoes the NIST test results.

Table 2

Key randomness test results table

Test items	Frequency	Block frequency	FFT	Serial	FWD	REV	Linear complexity
$p$	0.362	0.770	0.038	0.379,0.196	0.378	0.053	0.809

Some of the terms in Table 2 are explained as follows. FWD denotes Cumulative Sums (Cusum), FFT denotes Discrete Fourier Transform (Spectral) and REV denotes Random Excursions Variant.

As can be seen from Table 2, the p-values corresponding to the seven test terms are all greater than 0.01, indicating that the keys generated by this method have passed the NIST test and the generated keys have good randomness and availability.

In the experiment of extracting beacon frame interval, the window width is set to 100, the value of frame shift is set to 1, and the number of sampling points is set to the interval from 0 to 20000 in order to ensure the high accuracy of the energy profile. Under the IEEE 802.11 standard, the wireless AP broadcasts beacon frames with a period of 100 ms. The wireless signal is normalized to the interval [ $-$ 1, 1] and its energy curve is derived using the short-time energy method. The curves of the wireless signal amplitude and its short-time energy with the number of sampling points are shown in Fig. 5, and we can see that the short-time energy method has good results for the detection of wireless frames.

Figure 5.

Wireless signal amplitude and its short-time energy curve.

Set the threshold T to 0.2, find the mutation points, and separate the wireless frames and then calculate the duration of each wireless frame. After the wireless frames are separated, the set of beacon frame intervals can be derived according to their starting positions. The effect is shown in Fig. 6.

Figure 6.

The effect of separating the wireless frames according to the threshold value.

As can be seen from Fig. 6, the beacon frame extraction algorithm proposed in this paper is effective for separating beacon frames. On this basis, the inter-frame interval characteristics can be further obtained, and the identification of individual devices can be carried out on the basis of data encryption to ensure the security of data in the transmission process.

4. Conclusion

In this paper, we study the problem of secure transmission at the side end in the power IoT. In order to solve the problems of low key consistency and low key generation rate in the current physical layer key generation technology, we propose a reliable solution for secure transmission at the side end of the physical layer. First, the key consistency is improved by dynamically adjusting the guide frequency training sequence during the channel probing process; second, a multi-round iterative quantization method is proposed to improve the key generation rate. Finally, while data encryption is investigated, propose a method of frame interval extraction for wireless links to further realize real-time monitoring of third-party devices in wireless links. Simulation results show that the proposed method improves channel reciprocity and key consistency on the basis of guaranteed key randomness, improves the utilization of channel feature measurements and key generation rate, and at the same time can accurately extract the wireless frame interval, which provides a good foundation for third-party individual device identification, provides security for side-end data transmission in the power IoT, and provides interconnection and sharing of the underlying data of the new power system Lays the foundation of trust.

Footnotes

Acknowledgments

The authors acknowledge the State Grid Shanxi Electric Power Company Science and Technology Project Research (No. 52053022000D).

References

Yuan

Jiang

Aiqun

Guo

. Wireless Channel Key Generation for Multi-User Access Scenarios. IEEE/CIC International Conference on Communications in China (ICCC). 2022. pp. 179-184.

Yue

Feng

Wenjun

Jiaru

Shuguang

. Wireless traffic prediction with scalable gaussian process: Framework, algorithms, and verification. IEEE Journal on Selected Areas in Communications. 2019; 37(6): 1291-1306.

Yongjian

Linning

Shen

. Remote secure communication system based on wireless channel key generations. IEEE 6th International Conference on Signal and Image Processing (ICSIP). 2021. pp. 69-74.

Qiuhua

Mingyang

Guohua

Yizhi

Chunhua

. A Practical Secret Key Generation Scheme Based on Wireless Channel Characteristics for 5G Networks. IEICE Transactions on Information and Systems. 2020; 103(2): 230-238.

Tan

Huang

Liu

. A Method for Detecting Amplitude-Phase Joint Characteristic Parameters of Wireless Channel for Generating Key Parameters. Complexity. 2021.

Mingming

Chao

Meng

Fei

Ran

. Research on consistency enhancement of wireless channel key against RF fingerprint. Journal of Physics: Conference Series. 2020.

Altun

Basaran

Kurt

Ozdemir

. Scalable secret key generation for wireless sensor networks. IEEE Systems Journal. 2022; 16(4): 6031-6041.

Nguyen

Georgiou

Suppakitpaisarn

. Improved Localization Accuracy Using Machine Learning: Predicting and Refining RSS Measurements. IEEE Globecom Workshops (GC Wkshps). 2018. pp. 1-7.

Hongbo

Jie

Yan

Yingying

. Collaborative secret key extraction leveraging received signal strength in mobile wireless networks. Proc.31st IEEE Int.Conf.on Computer Communications. 2012. pp. 927-935.

10.

Xiao

Guo

Huang

. High-rate secret key generation aided by multiple relays for Internet of things. Electronics Letters. 2017; 53(17): 1198-1200.

11.

Chen

Kim

Vishwanath

. On the Key Generation Rate of Physically Unclonable Functions. 2018.

12.

Xiaoping

Waqas

Shanshan

. Power maximisation technique for generating secret keys by exploiting physical layer security in wireless communication. IET Communications. 2020; 14(5): 872-879.

13.

Yue

Feng

Wenjun

Chia-Han

Jiaru

Shuguang

. Scalable learning paradigms for data-driven wireless communication. IEEE Communications Magazine. 2020; 58(10): 81-87.

14.

Fang

Markwood

Liu

. Wireless-assisted key establishment leveraging channel manipulation. IEEE Transactions on Mobile Computing. 2021; 20(1): 263-275.

15.

Wang

Jiao

Yan

Zeng

Yang

. Security analysis of triangle channel-based physical layer key generation in wireless backscatter communications. IEEE Transactions on Information Forensics and Security. 2023; 18: 948-964.

16.

Denno

Hotta

Hou

. Maximum doppler frequency detection based on likelihood estimation with theoretical thresholds. IEICE Transactions on Communications. 2022; 105(5).

17.

Usman

Althunibat

Qaraqe

. A Channel State Information-Based Key Generation Scheme for Internet of Things. Security and Communication Networks. 2022.

18.

Yang

Zhang

Liu

. Fast and Secure Key Generation with Channel Obfuscation in Slowly Varying Environments. Proceedings – IEEE INFOCOM 2022 – IEEE Conference on Computer Communications. 2022. pp. 1-10.

19.

Chowdhury

Selouani

O’Shaughnessy

. Bayesian on-line spectral change point detection: a soft computing approach for on-line ASR. Int J Speech Technol 15. 2012; 5-23.

20.

Ding

Guan

Yang

Wang

. Performance analysis of wireless key generation with multi-bit quantization under imperfect channel estimation condition. Journal of Computer Applications. 2020; 40(1): 143-147.

21.

Dong

Liu

Zhou

Cai

. Voice activity detection based on short-time energy and noise spectrum adaptation. 6th International Conference on Signal Processing. Vol. 1. 2002. pp. 464-467.

22.

Limmanee

. Improper Quantization Decision and Key Entropy Loss in Wireless Physical-Layer Security with False Channel Statistics Knowledge. 2022 IEEE 33rd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC). 2022. pp. 1139-1145.