A trust management method suitable for internet of things perception nodes

Abstract

The application relates to a trust management method suitable for an Internet of Things sensing node. The Internet of Things sensing node network is a complex network composed of numerous heterogeneous nodes. In order to save energy or protect privacy and the like, the nodes tend to request to obtain trust information of other nodes, and refuse or ignore the request of other nodes to the nodes, so as to maximize the interests of the nodes. The utility of the whole system will be greatly reduced, and even the nodes will be influenced by malicious nodes to make wrong decisions. The application proves the existence of the evolutionary stable state through evolutionary game modeling on the trust management process, and through evolutionary dynamics analysis on the proposed game model, proves that the overall nodes will evolve to the direction of selecting the cooperation strategy in the state, and finally reach the state that most of the overall nodes select the cooperation strategy, so that the effectiveness of the scheme is guaranteed. The application effectively improves the trust management efficiency.

Description

Technical Field

[0001] This invention addresses trust management and game theory-related issues in IoT sensing node networks, specifically by using evolutionary game theory to establish relevant trust management mechanisms and improve trust management efficiency. Background Technology

[0002] In traditional IoT environments, devices collaborate and interact with resources through identity authentication. However, even authenticated malicious nodes within the IoT network can threaten secure cooperation. Furthermore, the complexity of IoT devices and the significant differences in computing resources and storage capabilities among them increase the difficulty of applying traditional security authentication and encryption technologies. Adding to this the differences in business attributes and trust characteristics among devices, the trust relationships within the entire IoT become even more complex. Trust mechanisms, as one of the effective methods for defending against internal network attacks and identifying malicious nodes, assess the trustworthiness of nodes within the network by combining different trust evidence and predicting the probability and capability of node cooperation. This allows for the filtering of malicious nodes before cooperation is established between nodes, thereby ensuring network security.

[0003] Currently, research on trust mechanisms mainly falls into two categories: policy-based trust models and reputation-based trust models. Policy-based trust models primarily define node trustworthiness strictly through digital certificates, rely on logical programming for node authorization, and typically depend on a centralized, authoritative trust management institution. Reputation-based trust models, on the other hand, typically assess node trustworthiness based on interaction experience between nodes, offering greater flexibility and scalability.

[0004] Game theory is the theory that primarily studies the strategies and decision-making of rational participants. Marked by the publication of John von Neumann and George Morgenstein's *Theory of Games and Economic Behavior* in 1944, game theory officially became a science. As a branch of applied mathematics, game theory is gradually being adopted as a general-purpose tool in other fields. Compared to traditional game theory, where participants are perfectly rational and the basic principle for choosing strategies is to maximize their own interests and accurately select the optimal strategy, evolutionary game theory, as a new branch of game theory, allows participants, based on bounded rationality, to continuously reflect on and adjust their strategies based on experience, enabling the overall system to maximize its interests when it tends towards stability. Summary of the Invention

[0005] IoT sensing nodes often need to cooperate and interact with each other to complete a series of tasks. The overall architecture is shown in the attached figure. Figure 1Trust management is an important mechanism to ensure the security of inter-node communication authentication. Currently, there has been extensive research on the acquisition of trust features in trust management. However, in practical applications, nodes may refuse to respond to the interaction requests of other nodes due to factors such as energy saving, resulting in low trust management efficiency and affecting system execution efficiency.

[0006] To address these issues, the inventors hope to introduce appropriate incentive mechanisms based on trust management to prevent selfish behavior by nodes.

[0007] To address the inefficiency of trust management mechanisms caused by selfish behavior of nodes, the inventors believe that the benefits gained from different strategic behaviors adopted by nodes during interaction should be modeled, and relevant incentive mechanisms should be established to encourage nodes to cooperate, thereby improving the efficiency of trust management mechanisms.

[0008] Based on evolutionary game theory and a dynamic evolutionary method using replicators, the inventors modeled the trust management behavior of IoT sensing node networks in a game-theoretic manner. First, they modeled the interaction behavior of the nodes, quantifying the interaction process and the gains obtained by each node, forming a payoff matrix for the two-way game. On this basis, a reputation-based reward and punishment mechanism was introduced. Interacting nodes accumulate reputation based on different behaviors, and nodes also change their strategies based on the reputation of their counterparts. Next, a reflection mechanism was introduced into the evolutionary game, where nodes adjust their strategies based on their own payoff levels to achieve higher returns. Subsequently, the evolutionary game model was analyzed using a dynamic evolutionary method, proving that the evolutionary game has an evolutionary stable state. Based on this stable state, the incentive mechanism in trust management was dynamically adjusted, enabling the IoT sensing nodes as a whole to evolve towards a cooperative strategy.

[0009] By modeling the trust management process using evolutionary game theory and introducing reputation and reflection mechanisms, the inventors implemented a corresponding solution. Evolutionary dynamics analysis of the proposed game model proved the existence of an evolutionary stable state. In this state, all nodes evolve towards choosing a cooperative strategy, ultimately reaching a state where the majority of nodes choose a cooperative strategy, thus ensuring the effectiveness of the solution.

[0010] A trust management method suitable for IoT sensing nodes, characterized in that:

[0011] Before performing game theory modeling on the node network, the following settings are first made for the perception node network:

[0012] 1) The entire perception node network consists of N rational perception nodes. Trust management between nodes includes direct trust measurement and indirect trust measurement. Direct measurement obtains trust information by interacting with the target node, while indirect trust measurement obtains trust information by interacting with the target node's neighbor nodes.

[0013] 2) Sending a message requires the same energy c for all nodes, while the energy consumption for receiving messages is negligible; the basic game theory model for interaction between nodes is defined below;

[0014] Process 1: Define the game process of the nodes as a triple, G = (P, S, u), where P is the set of game participants, i.e., the sensing nodes in the network, P = {i, j}, i, j ∈ N, S is the set of strategies adopted by each node, S = {s1, s2} = {cooperation (C), non-cooperation (UN)}, and u is the payoff function of the node.

[0015] Define the payoff function for node i as follows:

[0016] u i (S i ,S j ) = S j b i -S i (cn i +p) (1)

[0017] Among them, S i ,S j These are the game strategies chosen by node i and node j in time slot t, respectively; b i When the target node chooses a cooperation strategy, b is the benefit of node i. i =δe a Where δ is the maximum benefit obtained by node i when cooperating with the target node, and a is the influence factor of this trust information on node i's trust decision. The larger a is, the greater the impact of this trust acquisition on the overall trust decision. 0 ≤ a < 1, default a = 0.5; n i is the number of messages sent by node i in a single trust acquisition; p is the loss incurred by node i due to privacy disclosure after choosing a cooperation strategy;

[0018] The payoff function for node j is

[0019] u j (S j ,S i ) = S i b j -S j (cn j +p) (2)

[0020] Process 2: Based on the payoff function of the utility node, the payoff matrices of both players can be determined; if node i chooses a cooperative strategy, it will honestly respond to the requests of other nodes, at which point S i =1, if a node acts selfishly due to energy saving or other considerations, i.e., ignores messages from other nodes, then S i=0; From this, we can obtain the benefit matrix of the trust management model;

[0021] As can be seen from the payoff matrix, when both sides choose not to cooperate, the game reaches a Nash equilibrium. At this point, no matter what decision the other side makes, the payoff of the player will be higher than when they cooperate. At this point, the payoff of both nodes is 0. Trust management under this situation will fail to obtain trust evidence from the other side due to the non-cooperative behavior of the nodes. The nodes must then choose other nodes to repeat the trust management work, which greatly reduces the efficiency of trust management.

[0022] Incentive mechanism:

[0023] In the process of trust management, a reputation incentive mechanism is introduced. This mechanism rewards nodes that choose a cooperative strategy with reputation and deducts reputation from nodes that choose a non-cooperative strategy. The method also allows each node to reflect on and adjust its strategy based on its reputation and benefits, thus forming an evolutionary game.

[0024] Process 2.1: Let R be the reputation value of node i at time t in the sensing node network. i If (t), then when a node chooses a cooperation strategy, the node's reputation value is defined as...

[0025] R i (t)=R i (t')×λ(t,t')+r i (3)

[0026] If a node chooses a non-cooperative strategy, its reputation value is defined as follows:

[0027] R i (t)=R i (t')×λ(t,t')-r i (4)

[0028] Where t' is the time of the last interaction, and r i Let λ(t,t') be the base value for accumulated reputation, and let λ(t,t') be the time decay function. In this paper, let... Where c is the adjustment factor of the decay function, to prevent the decay function from being too small, which would reduce the impact of accumulated reputation. The larger c is, the slower the overall reputation of the node decays. The default c = 0.5. Under the action of the decay function, the reputation value previously accumulated by the node will decay over time, and the longer the time since the current interaction, the greater the degree of decay.

[0029] Process 2.2: When IoT sensing nodes engage in trust games, they adjust their strategies based on the other party's reputation score. Nodes with lower reputation scores are more likely to be rejected. The adjustment probability is defined as follows: Where R i(t) represents the reputation of node i at time t, and N represents the total reputation of the nodes. According to formulas (1) and (2) above, in the trust game process, the expected payoff of node j for the trust game against node i is:

[0030] 1) If node j chooses a non-cooperative strategy, it will adjust to a cooperative strategy based on the other party's reputation value. Then, node i's expected return is...

[0031] U i (S i ,S j )=(1-P)u i (S i ,UN)+Pu i (S i (C) (5)

[0032] 2) If node j chooses a cooperative strategy, it will adjust to a non-cooperative strategy due to the other party's reputation value. Then, node i's expected return is...

[0033] U i (S i ,S j )=(1-P)u i (S i ,C)+Pu i (S i (6)

[0034] Process 2.3: Based on the accumulation of reputation, a reflection mechanism is introduced to measure the individual's strategy adjustment in the evolutionary process, making the evolutionary process more accurate;

[0035] In a trust game involving sensing nodes, the strategy s is chosen. i The average reflection rate of nodes ∈S For mu(s) i ,S)+nR(t) where mu(s) i ,S)+nR(t)>0 and are constants, where m=0.5, n=0.5, u(s) i S) represents the strategy s to choose when the overall situation is S. i Let R(t) be the expected total revenue of a node, and R(t) be the node reputation.

[0036] Process 2.4: In the trust game of sensing nodes, choosing strategy s j When a node ∈S reflects, its strategy is adjusted to s. i probability for Where g is a constant such that g + u(s) h If S > 0, then g = 0.5 by default. These represent the overall node selection strategy. s h The proportion, u(s) i ,S),u(s h S and S represent the node selection strategies under the overall situation of S. s h Expected returns;

[0037] In a finite population, during trust games, due to the existence of a reflection mechanism, different strategies of the perceived nodes will flow into and out of each other, thus continuously evolving; for the choice of strategy s i For the nodes, the total outflow is Total inflow is in These represent the node selection strategies s, respectively. i ,s j proportion, These represent the selection strategies s, respectively. i ,s j The proportion of reflection at each node;

[0038] Evolutionary dynamics analysis:

[0039] Since the evolutionary stable state x* in an evolutionary game is stable—meaning that even if a few mutations occur in the population under this state, it will eventually return to the x* state—according to the stability property of the equilibrium point of the differential equation, the evolutionary stable state x* must satisfy the following conditions: the function value of the dynamic equation at x* is equal to 0, and the derivative at x* is less than 0, i.e., F(x*) = 0 and F'(x*) < 0; let and These represent the proportions of nodes choosing cooperative and non-cooperative strategies, respectively, and satisfy the following conditions: From the dynamic equation of the replicon The dynamic equation for the proportion of individuals choosing cooperative strategies is obtained;

[0040] Step 3.1: Substituting the payoff function and incentive mechanism of the evolutionary game model into the evolutionary dynamics analysis, we can obtain the expected payoff of the node that chooses the cooperative strategy as follows:

[0041]

[0042] Where U(s1,s1) and U(s1,s2) represent the payoffs of the opposing node when the node chooses a cooperation strategy and when the opposing node chooses a rejection strategy, respectively. These represent the probabilities of a node adjusting to a cooperative strategy and a rejection strategy, respectively.

[0043] The expected return of a node that chooses the non-cooperative strategy is

[0044]

[0045] Where U(s2,s1) and U(s2,s2) represent the benefits of the opposing node when the node chooses the rejection strategy and when the opposing node chooses the cooperation and rejection strategies, respectively.

[0046] According to formulas (7) and (8), the average profit of the entire network is:

[0047]

[0048] From the replier dynamic equation, we can see that the replication dynamic equation for the trust evolution of the cooperative strategy is:

[0049]

[0050] Substituting equation (9) into equation (10) yields

[0051]

[0052] Process 3.2: Equation (11) is the dynamic evolution equation of the evolutionary game model based on reputation reflection. It reflects the evolutionary trend of the entire evolutionary group's strategy. According to the characteristics of evolutionary stability, the whole in the evolutionary stable state will quickly return to the stable state even if there is a small mutation. That is to say, in the process of IoT trust management, if a small number of nodes in the IoT environment in the evolutionary stable state choose a non-cooperative strategy, they will quickly adjust to a cooperative strategy due to the overall evolutionary mechanism. The following is an analysis of its evolutionary stable state.

[0053] From equation (11), we can see that And let Solving this equation reveals that the evolutionary game has two stable states.

[0054] When u(s1,x)-u(s2,x)<0, This is the stable state of the evolutionary game based on reputation reflection in the Internet of Things; at this time, the Nash equilibrium of the evolutionary game is that both sides choose the strategy of non-cooperation, which is strictly dominant and evolutionarily stable; in this state, no matter what strategy node j chooses, the payoff of node i choosing the strategy of non-cooperation is always greater than the payoff of choosing the strategy of cooperation; regardless of the overall state, after evolution, it will eventually reach a stable state in which all nodes choose the strategy of non-cooperation.

[0055] When u(s1,x)-u(s2,x)>0, This is the stable state of the evolutionary game based on reputation reflection in the Internet of Things; at this point, the Nash equilibrium of the evolutionary game is a strategy in which both parties choose to cooperate, and this equilibrium is strictly dominant and evolutionarily stable.

[0056] Trust management mechanism:

[0057] The entire system needs to satisfy the condition u(s1,x)-u(s2,x)>0 as much as possible; substituting formulas (7) and (8) into them yields...

[0058]

[0059] In summary, by adjusting the energy consumption c of node information transmission and the probability P of node policy adjustment based on reputation, we can make u(s1,x)-u(s2,x)>0 in the entire system, thus causing the nodes in the system to evolve towards a stable state of choosing a cooperative policy.

[0060] After testing, it was found that the technical solution proposed by the inventor can effectively reduce the proportion of selfish nodes in the trust management process of IoT sensing node network, and effectively improve the efficiency of trust management. Attached Figure Description

[0061] Figure 1 IoT sensing node network architecture

[0062] Figure 2 Payoff matrix of trust management game model Detailed Implementation

[0063] Trust management methods include game theory models, incentive mechanisms, evolutionary dynamics analysis, and trust management mechanism analysis.

[0064] Game theory model:

[0065] Before performing game theory modeling on the node network, we first make the following assumptions about the perceptual node network.

[0066] 3) The entire perception node network consists of N rational perception nodes. Trust management between nodes includes direct trust measurement and indirect trust measurement. Direct measurement obtains trust information by interacting with the target node, while indirect trust measurement obtains trust information by interacting with the target node's neighbor nodes.

[0067] 4) All nodes need to consume the same amount of energy c to send a message, while the energy consumption for receiving messages is negligible.

[0068] The basic game theory model for interaction between nodes is defined below.

[0069] Process 1: The game process of the nodes can be defined as a triple, G = (P, S, u), where P is the set of game participants, i.e., the sensing nodes in the network, P = {i, j}, i, j ∈ N, S is the set of strategies that each node can take, S = {s1, s2} = {cooperation (C), non-cooperation (UN)}, and u is the payoff function of the node.

[0070] Define the payoff function for node i as follows:

[0071] u i (S i ,S j ) = S j b i -S i (cn i +p) (1)

[0072] Among them, S i ,S j These are the game strategies chosen by node i and node j in time slot t, respectively; b i When the target node chooses a cooperation strategy, b is the benefit of node i. i =δe a Where δ is the maximum benefit that node i can obtain when cooperating with the target node, and a is the influence factor of this trust information on node i's trust decision. The larger a is, the greater the impact of this trust acquisition on the overall trust decision. 0 ≤ a < 1, default a = 0.5; n i is the number of messages sent by node i in a single trust acquisition; p is the loss incurred by node i due to privacy disclosure after choosing a cooperation strategy.

[0073] Similarly, the payoff function for node j is

[0074] u j (S j ,S i ) = S i b j -S j (cn j +p) (2)

[0075] Step 2: Based on the payoff function of the utility node, the payoff matrices of both players can be determined. If node i chooses a cooperative strategy, it will honestly respond to the requests of other nodes, at which point S... i =1, if a node acts selfishly due to energy saving or other considerations, i.e., ignores messages from other nodes, then S i =0. Therefore, the payoff matrix of the trust management model can be obtained, as shown in the appendix. Figure 2 .

[0076] The payoff matrix shows that when both sides choose not to cooperate, the game reaches a Nash equilibrium. At this point, regardless of the other side's decision, one's own payoff will be higher than when cooperating. In this situation, the payoffs for both nodes are zero. Trust management under these circumstances will fail because the nodes' non-cooperative behavior prevents them from obtaining evidence of the other party's trust. Nodes must then choose other nodes to repeat the trust management process, significantly reducing its efficiency.

[0077] Incentive mechanism:

[0078] In the process of trust management, a reputation incentive mechanism is introduced. This mechanism rewards nodes that choose a cooperative strategy with reputation and deducts reputation from nodes that choose a non-cooperative strategy. This allows each node to reflect on and adjust its strategy based on its reputation and benefits, thus forming an evolutionary game.

[0079] Process 1: Let R be the reputation value of node i at time t in the sensing node network. i If (t), then when a node chooses a cooperation strategy, the node's reputation value is defined as...

[0080] R i (t)=R i (t')×λ(t,t')+r i (3)

[0081] If a node chooses a non-cooperative strategy, its reputation value is defined as follows:

[0082] R i (t)=R i (t')×λ(t,t')-r i (4)

[0083] Where t' is the time of the last interaction, and r i Let λ(t,t') be the base value for accumulated reputation, and let λ(t,t') be the time decay function. In this paper, let... Here, 'c' is an adjustment factor for the decay function, preventing it from becoming too small and diminishing the impact of accumulated reputation. A larger 'c' results in slower overall reputation decay for the node; the default value is 0.5. Under the influence of the decay function, the previously accumulated reputation value of a node decays over time, with the decay rate increasing the time elapsed since the last interaction.

[0084] Process 2: When IoT sensing nodes engage in trust games, they adjust their strategies based on the other party's reputation score. Nodes with lower reputation scores are more likely to be rejected. The adjustment probability is defined as follows: Where R i (t) represents the reputation of node i at time t, and N represents the total reputation of the nodes. According to formulas (1) and (2) above, in the trust game process, the expected payoff of node j for the trust game against node i is:

[0085] 3) If node j chooses a non-cooperative strategy, it will adjust to a cooperative strategy based on the other party's reputation value. Then, node i's expected return is...

[0086] U i (S i ,S j )=(1-P)u i (S i ,UN)+Pu i (S i(C) (5)

[0087] 4) If node j chooses a cooperative strategy, it will adjust to a non-cooperative strategy due to the other party's reputation value. Then, node i's expected return is...

[0088] U i (S i ,S j )=(1-P)u i (S i ,C)+Pu i (S i (6)

[0089] Process 3: Based on the accumulation of reputation, a reflection mechanism is introduced to measure the individual's strategy adjustments during the evolutionary process, making the evolutionary process more accurate. During the trust game, nodes will reflect on their strategies at certain time intervals, and each reflection has a certain probability of leading to a strategy adjustment.

[0090] In a trust game involving sensing nodes, the strategy s is chosen. i The average reflection rate of nodes ∈S For mu(s) i ,S)+nR(t) where mu(s) i ,S)+nR(t)>0 and are constants, where m=0.5, n=0.5, u(s) i S) represents the strategy s to choose when the overall situation is S. i Let R(t) be the expected total revenue of a node, and let R(t) be the node reputation.

[0091] Process 4: In the trust game of sensing nodes, choosing strategy s j When a node ∈S reflects, its strategy is adjusted to s. i probability for Where g is a constant such that g + u(s) h If S > 0, then g = 0.5 by default. These represent the overall node selection strategy. s h The proportion, u(s) i ,S),u(s h S and S represent the node selection strategies under the overall situation of S. s h Expected returns.

[0092] In a finite population, during trust games, perceptual nodes, due to the existence of a reflection mechanism, will experience mutual inflows and outflows between populations with different strategies, thus continuously evolving. For the choice of strategy s... i For the nodes, the total outflow is Total inflow is in These represent the node selection strategies s, respectively. i ,s j proportion, These represent the selection strategies s, respectively. i ,s j The proportion of reflection at each node.

[0093] Evolutionary dynamics analysis:

[0094] Since the evolutionary stable state x* in an evolutionary game is stable—meaning that even if a few mutations occur in the population under this state, it will eventually return to the x* state—according to the stability property of the equilibrium point of the differential equation, the evolutionary stable state x* must satisfy the following conditions: the function value of the dynamic equation at x* is equal to 0, and the derivative at x* is less than 0, i.e., F(x*) = 0 and F'(x*) < 0. Let... and These represent the proportions of nodes choosing cooperative and non-cooperative strategies, respectively, and satisfy the following conditions: From the dynamic equation of the replicon The dynamic equation for the proportion of individuals choosing a cooperative strategy can be obtained.

[0095] Process 1: Substituting the payoff function and incentive mechanism of the evolutionary game model into the evolutionary dynamics analysis, we can obtain the expected payoff of the node that chooses the cooperative strategy as follows:

[0096]

[0097] Where U(s1,s1) and U(s1,s2) represent the payoffs of the opposing node when the node chooses a cooperation strategy and when the opposing node chooses a rejection strategy, respectively. These represent the probabilities of a node adjusting to a cooperative strategy and a rejection strategy, respectively.

[0098] The expected return of a node that chooses the non-cooperative strategy is

[0099]

[0100] Where U(s2,s1) and U(s2,s2) represent the benefits of the opposing node when it chooses the cooperation and rejection strategies, respectively, when the node chooses the rejection strategy.

[0101] According to formulas (7) and (8), the average profit of the entire network is:

[0102]

[0103] From the replier dynamic equation, we can see that the replication dynamic equation for the trust evolution of the cooperative strategy is:

[0104]

[0105] Substituting equation (9) into equation (10) yields

[0106]

[0107] Process 2: Equation (11) is the dynamic evolution equation of the evolutionary game model based on reputation reflection, which reflects the evolutionary trend of the entire evolutionary group's strategy. According to the characteristics of evolutionary stability, the whole in an evolutionary stable state can quickly return to a stable state even if a small mutation occurs. That is to say, in the process of IoT trust management, if a small number of nodes in an IoT environment in an evolutionary stable state choose a non-cooperative strategy, they will quickly adjust to a cooperative strategy due to the overall evolutionary mechanism. The following is an analysis of its evolutionary stable state.

[0108] From equation (11), we can see that And let Solving this equation reveals that the evolutionary game has two stable states.

[0109] When u(s1,x)-u(s2,x)<0, This represents the stable state of an evolutionary game based on reputation reflection in the Internet of Things (IoT). At this point, the Nash equilibrium of the evolutionary game is that both nodes choose a non-cooperative strategy; this equilibrium is strictly dominant and evolutionarily stable. In this state, regardless of the strategy chosen by node j, the payoff for node i to choose a non-cooperative strategy is always greater than the payoff for it to choose a cooperative strategy. Regardless of the overall state, through evolution, a stable state will eventually be reached where all nodes choose a non-cooperative strategy.

[0110] When u(s1,x)-u(s2,x)>0, This represents the stable state of an evolutionary game based on reputation reflection in the Internet of Things (IoT). At this point, the Nash equilibrium of the evolutionary game is a strategy where both nodes choose cooperation; this equilibrium is strictly dominant and evolutionarily stable. In this state, regardless of the strategy chosen by node j, the payoff for node i to choose the cooperation strategy is always greater than its payoff for choosing the cooperation strategy. Regardless of the overall state, through evolution, a stable state will eventually be reached where all nodes choose the cooperation strategy.

[0111] Trust management mechanism:

[0112] In the Internet of Things (IoT) sensing node network, when nodes perform trust management, in order to ensure high management efficiency, it is necessary to enable rational nodes to actively participate in the interaction process of trust evidence as much as possible, that is, to avoid the non-cooperative strategy of nodes. Therefore, it is necessary to make all nodes in the entire sensing node network evolve towards a stable state of choosing a cooperative strategy. Thus, the entire system needs to satisfy the condition u(s1,x)-u(s2,x)>0 as much as possible.

[0113] Substituting formulas (7) and (8) into the equations, we can obtain...

[0114]

[0115] In summary, by adjusting the energy consumption c of node information transmission and the probability P of node policy adjustment based on reputation, making u(s1,x)-u(s2,x)>0 in the entire system, the nodes of the system as a whole can evolve towards a stable state of choosing a cooperative policy.

Claims

1. A trust management method suitable for IoT sensing nodes, characterized in that: Before performing game theory modeling on the node network, the following settings are first made for the perception node network: 1) The entire perception node network consists of N rational perception nodes. Trust management between nodes includes direct trust measurement and indirect trust measurement. Direct trust measurement obtains trust information by interacting with the target node, while indirect trust measurement obtains trust information by interacting with the target node's neighbor nodes. 2) Sending a message requires the same amount of energy from all nodes, while the energy consumption for receiving messages is negligible; The basic game theory model for interaction between nodes is defined below; Process 1: Define the game process of the nodes as a triple, G = (P1, S, u), where P1 is the set of game participants, i.e., the sensing nodes in the network, P1 = {i, j}, i, j ∈ N, S is the set of strategies adopted by each node, S = {s1, s2} = {cooperation (C), non-cooperation (UN)}, and u is the payoff function of the node; Define the payoff function for node i as follows: you i (S i ,S j )=S j b i -S i (cn i +p) (1) Among them, S i ,S j These are the game strategies chosen by node i and node j in time slot t, respectively; b i When the target node chooses a cooperation strategy, b is the benefit of node i. i =δe a Where δ is the maximum benefit obtained by node i when cooperating with the target node, and a is the influence factor of this trust information on node i's trust decision. The larger a is, the greater the impact of this trust acquisition on the overall trust decision. 0 ≤ a < 1, default a = 0.5; n i is the number of messages sent by node i in a single trust acquisition; p is the loss incurred by node i due to privacy disclosure after choosing a cooperation strategy; The payoff function for node j is you j (S j ,S i )=S i b j -S j (cn j +p) (2) Process 2: Based on the payoff function of the utility node, the payoff matrices of both players can be determined; if node i chooses a cooperative strategy, it will honestly respond to the requests of other nodes, at which point S i =1, if a node acts selfishly due to energy saving or other considerations, i.e., ignores messages from other nodes, then S i =0; From this, we can obtain the benefit matrix of the trust management model; As can be seen from the payoff matrix, when both sides choose not to cooperate, the game reaches a Nash equilibrium. At this point, no matter what decision the other side makes, the payoff of the player will be higher than when they cooperate. At this point, the payoff of both nodes is 0. Trust management under this situation will fail to obtain trust evidence from the other side due to the non-cooperative behavior of the nodes. The nodes must then choose other nodes to repeat the trust management work, which greatly reduces the efficiency of trust management. Incentive mechanism: In the process of trust management, a reputation incentive mechanism is introduced. This mechanism rewards nodes that choose a cooperative strategy with reputation and deducts reputation from nodes that choose a non-cooperative strategy. The method also allows each node to reflect on and adjust its strategy based on its reputation and benefits, thus forming an evolutionary game. Process 2.1: Let R be the reputation value of node i at time t in the sensing node network. i If (t), then when a node chooses a cooperation strategy, the node's reputation value is defined as... R i (t)=R i (t')×λ(t,t')+r i (3) If a node chooses a non-cooperative strategy, its reputation value is defined as follows: R i (t)=R i (t')×λ(t,t')-r i (4) Where t' is the time of the last interaction, and r i Let λ(t,t') be the base value for accumulated reputation, and let λ(t,t') be the time decay function. In this paper, let... Where c1 is the adjustment factor of the decay function to prevent the decay function from being too small, thus reducing the impact of accumulated reputation. The larger c1 is, the slower the overall reputation decay of the node. The default c1 = 0.

5. Under the action of the decay function, the previously accumulated reputation value of the node will decay over time, and the longer the time since the current interaction, the greater the decay. Process 2.2: When IoT sensing nodes engage in trust games, they will adjust their strategies according to the other party's reputation value. Nodes with low reputation are more likely to be rejected. The adjustment probability is defined as... Where R i (t) represents the reputation of node i at time t, and N represents the total number of nodes. According to formulas (1) and (2) above, in the trust game process, the expected payoff of node j against node i is: If node j chooses the non-cooperative strategy, it will adjust to the cooperative strategy due to the reputation value of the other party, then the expected payoff of node i is: The i (S i ,S j )=(1-P)u i (S i ,A)+Pu i (S i ,C) (5) 1) If node j chooses a cooperative strategy, it will adjust to a non-cooperative strategy due to the other party's reputation value. Then, node i's expected return is... The i (S i ,S j )=(1-P)u i (S i ,C)+Pu i (S i (6) Process 2.3: Based on the accumulation of reputation, a reflection mechanism is introduced to measure the individual's strategy adjustment in the evolutionary process, making the evolutionary process more accurate; In a trust game involving sensing nodes, the strategy s is chosen. i Nodes ∈S undergo reflection; the average reflection rate is... For mu(s) i ,S)+nR(t), where mu(s) i ,S)+nR(t)>0 and are constants, where m=0.5, n=0.5, u(s) i S) represents the strategy s to choose when the overall situation is S. i Let R(t) be the expected total revenue of a node, and R(t) be the node reputation. Process 2.4: In the trust game of sensing nodes, choosing strategy s j When a node ∈S reflects, its strategy is adjusted to s. i probability for Where g is a constant such that g + u(s) h If S > 0, then g = 0.5 by default. These represent the overall node selection strategy s. i s h The proportion, u(s) i ,S),u(s h S and S represent the node selection strategies s and s respectively, given an overall situation of S. i s h Expected returns; In a finite population, during trust games, due to the existence of a reflection mechanism, different strategies of the perceived nodes will flow into and out of each other, thus continuously evolving; for the choice of strategy s i For the nodes, the total outflow is Total inflow is in These represent the node selection strategies s, respectively. i ,s j proportion, These represent the selection strategies s, respectively. i ,s j The proportion of reflection at each node; Evolutionary dynamics analysis: Since the evolutionary stable state x* in an evolutionary game is stable—meaning that even if a few mutations occur in the population under this state, it will eventually return to the x* state—according to the stability property of the equilibrium point of the differential equation, the evolutionary stable state x* must satisfy the following conditions: the function value of the dynamic equation at x* is equal to 0, and the derivative at x* is less than 0, i.e., F(x*) = 0 and F'(x*) < 0; let and These represent the proportions of nodes choosing cooperative and non-cooperative strategies, respectively, and satisfy the following conditions: From the dynamic equation of the replicon The dynamic equation for the proportion of individuals choosing cooperative strategies is obtained; Step 3.1: Substituting the payoff function and incentive mechanism of the evolutionary game model into the evolutionary dynamics analysis, we can obtain the expected payoff of the node that chooses the cooperative strategy as follows: Where U(s1,s1) and U(s1,s2) represent the payoffs of the opposing node when the node chooses a cooperation strategy and when the opposing node chooses a rejection strategy, respectively. These represent the probabilities of a node adjusting to a cooperative strategy and a rejection strategy, respectively. The expected return of a node that chooses the non-cooperative strategy is Where U(s2,s1) and U(s2,s2) represent the benefits of the opposing node when the node chooses the rejection strategy and when the opposing node chooses the cooperation and rejection strategies, respectively. According to formulas (7) and (8), the average profit of the entire network is: From the replier dynamic equation, we can see that the replication dynamic equation for the trust evolution of the cooperative strategy is: Substituting equation (9) into equation (10) yields Process 3.2: Equation (11) is the dynamic evolution equation of the evolutionary game model based on reputation reflection. It reflects the evolutionary trend of the entire evolutionary group's strategy. According to the characteristics of evolutionary stability, the whole in the evolutionary stable state will quickly return to the stable state even if there is a small mutation. That is to say, in the process of IoT trust management, if a small number of nodes in the IoT environment in the evolutionary stable state choose a non-cooperative strategy, they will quickly adjust to a cooperative strategy due to the overall evolutionary mechanism. The following is an analysis of its evolutionary stable state. From equation (11), we can see that And let Solving this equation reveals that the evolutionary game has two stable states. When u(s1,x)-u(s2,x)<0, This is the stable state of the evolutionary game based on reputation reflection in the Internet of Things; at this time, the Nash equilibrium of the evolutionary game is that both sides choose the strategy of non-cooperation, which is strictly dominant and evolutionarily stable; in this state, no matter what strategy node j chooses, the payoff of node i choosing the strategy of non-cooperation is always greater than the payoff of choosing the strategy of cooperation; regardless of the overall state, after evolution, it will eventually reach a stable state in which all nodes choose the strategy of non-cooperation. When u(s1,x)-u(s2,x)>0, This is the stable state of the evolutionary game based on reputation reflection in the Internet of Things; at this point, the Nash equilibrium of the evolutionary game is a strategy in which both parties choose to cooperate, and this equilibrium is strictly dominant and evolutionarily stable. Trust management mechanism: The entire system needs to satisfy the condition u(s1,x)-u(s2,x)>0 as much as possible; substituting formulas (7) and (8) into them yields... In summary, by adjusting the energy consumption c of node information transmission and the probability P of node policy adjustment based on reputation, we can make u(s1,x)-u(s2,x)>0 in the entire system, thus causing the nodes in the system to evolve towards a stable state of choosing a cooperative policy.

Images