A method and system for array graph hopping of a cloud-native honeycomb array

Abstract

The application provides a kind of array graph hopping method and system for cloud native honeycomb, it is related to artificial intelligence technical field.The method provided by the application comprises: obtaining an attacker action set and constructing an array graph configuration set according to a technical component, and constructing an attack set and a defense set based on a repeated bayesian game model;Generate array graph hopping strategy based on attack set and defense set;Define the income set of attack set and defense set based on common vulnerability scoring system, and construct honeycomb target equation based on repeated bayesian game model, optimize array graph hopping strategy based on deep Q network algorithm.The application constructs different array graph configurations by different technical components, builds network penetration and array graph hopping based on repeated bayesian game modeling, optimizes array graph hopping strategy using deep Q network algorithm, effectively solves the array graph strategy optimization problem under large-scale configuration space and maximizes the expected utility of honeycomb, improves the computing efficiency while ensuring the dynamic adaptability and effectiveness of array graph strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a method and system for array graph hopping of a cloud-native honeycomb array. BACKGROUND

[0002] A four-honey system is a kind of active defense mechanism composed of four parts: honey point, honey courtyard, honey hole and honeycomb, and is applied to the field of network security. Among them, the honeycomb as the central control unit is responsible for overall scheduling of the entire honeycomb system, and dynamically adjusts according to network attack behavior to achieve the purpose of deception defense. Cloud-native applications decompose the application program into multiple independent and loosely coupled technology components, combined with different container runtimes or database management systems, to build multi-dimensional configuration capabilities for honeycomb array. However, in the actual network attack and defense scene, the honeycomb often cannot obtain all the information of the network attacker, and how to formulate the optimal honeycomb array strategy under incomplete information is a key challenge for honeycomb hopping technology. In addition, array hopping needs to consume certain resources, and how to minimize the hopping cost while achieving the optimal deception defense effect is also a problem that needs to be solved by honeycomb hopping technology.

[0003] At present, the existing research proposes a dynamic heterogeneous scheduling method, which builds a heterogeneous image resource pool, periodically selects heterogeneous containers for dynamic rotation and cleaning operation to improve the overall security of the container cloud environment. However, this method cannot meet the minimum defense cost while considering the optimal deception defense effect in the face of incomplete information environment. Therefore, there is an urgent need to provide a solution to improve the above problems. SUMMARY

[0004] The present application relates to the technical field of artificial intelligence, and in particular to a method and system for array graph hopping of a cloud-native honeycomb array.

[0005] In a first aspect, the present application provides a method for array graph hopping of a cloud-native honeycomb array, comprising:

[0006] obtaining an attack action set and constructing an array configuration set according to the technology components in the cloud-native technology stack, and constructing an attack set and a defense set based on a repeated Bayesian game model;

[0007] generating an array graph hopping strategy based on the attack set and the defense set;

[0008] optimizing the array graph hopping strategy based on a deep Q network algorithm.

[0009] The application provides a kind of array graph hopping method for cloud native honeycomb, by obtaining attacker action set from national vulnerability database and common vulnerability disclosure and constructing array graph configuration set according to technical components, then based on repeated bayesian game model, attack-defense process modeling is carried out, array graph hopping strategy is generated based on array graph configuration set, then based on general vulnerability scoring system, define the set of benefits and based on repeated bayesian game model, build honeycomb equation, finally based on deep Q network algorithm, optimize array graph hopping strategy, realize array graph strategy optimization problem under large-scale configuration space and maximize the expected utility of honeycomb, improve the computational efficiency while ensuring the dynamic adaptability and effectiveness of array graph strategy.

[0010] Optionally, the technical components include container runtime, cloud native network, database management system, container image library and orchestration and scheduling system.

[0011] Optionally, the attacker action set includes attack behavior against container runtime, attack behavior against container image library, attack behavior against network protocol stack, attack behavior against database management system and attack behavior against orchestration and scheduling system.

[0012] Optionally, when constructing array graph configuration set according to technical components, it includes: selecting at least one component from each of the three types of technical components, container runtime, cloud native network and database management system, to form different array graph configurations, and different array graph configurations ultimately constitute the array graph configuration set.

[0013] Optionally, the array graph hopping strategy is a decision-making process based on probability distribution to dynamically adjust the transition from the current array graph configuration to the next effective array graph configuration.

[0014] Optionally, when optimizing the array graph hopping strategy based on deep Q network algorithm, it includes: quantifying the benefits of the attack set and the defense set based on the general scoring system to generate the benefit set, which includes attack benefits and defense benefits, combining the benefit set and based on repeated bayesian game model, building honeycomb target equation, and based on deep Q network algorithm, solving the optimal array graph hopping strategy, the solving process is as follows:

[0015] Step 1, initialize attack-defense environment model parameters, experience replay buffer, evaluation network parameters and target network parameters, the attack-defense environment model parameters include array graph configuration set, attack strategy, defense strategy, attacker attack probability and benefit set;

[0016] Step 2, in each training cycle, initialize environment and state, and extract attacker and his attack action set according to the prior probability;

[0017] Step 3, selecting a graph action based on an ε-greedy policy, wherein the graph action is randomly selected with a probability of ε, otherwise the Q-value action considered optimal by the evaluation network is selected;

[0018] Step 4, calculating a corresponding reward based on the repeated Bayesian game model and the set of returns and storing training data into the experience replay buffer;

[0019] Step 5, randomly sampling a training sample from the experience replay buffer and calculating a Q-value of the current state based on the evaluation network and a target Q-value of the next state based on a target network;

[0020] Step 6, calculating a loss function and updating the evaluation network parameters;

[0021] Repeating steps 2 to 6 until the graph hopping strategy converges, and outputting an optimal graph hopping strategy.

[0022] In a second aspect, the present application further provides a graph hopping system for a cloud-native honeycomb, comprising:

[0023] A model construction module, which acquires a set of attacker actions and constructs a set of graph configurations according to technical components in a cloud-native technology stack, and respectively constructs an attack set and a defense set based on a repeated Bayesian game model;

[0024] A strategy generation module, which generates a graph hopping strategy based on the attack set and the defense set;

[0025] A strategy optimization module, which optimizes the graph hopping strategy based on a deep Q-network algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A flowchart of a graph hopping method for a cloud-native honeycomb provided by an embodiment of the present application.

[0027] Figure 2 A structural diagram of a graph hopping system for a cloud-native honeycomb provided by an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. Unless otherwise defined, the technical terms or scientific terms used herein should have the usual meanings understood by those skilled in the art.

[0029] Referring to Figure 1 The application provides a method for array graph hopping of a cloud native honeycomb, comprising the following steps:

[0030] S1, obtaining an attacker action set and constructing an array graph configuration set according to technical components in a cloud native technology stack, and respectively constructing an attack set and a defense set based on a repeated Bayesian game model;

[0031] S2, generating an array graph hopping strategy based on the attack set and the defense set;

[0032] S3, optimizing the array graph hopping strategy based on a deep Q network algorithm.

[0033] In fact, the array graph hopping method provided by the application can make the honeycomb adjust the array graph hopping strategy according to the latest information, realize the array graph strategy optimization problem in a large-scale configuration space, and maximize the expected utility of the honeycomb, by obtaining an attacker action set according to common vulnerability disclosure data recorded in a national vulnerability database, constructing an array graph configuration set according to technical components, then modeling the attack and defense process based on a repeated Bayesian game model, generating an array graph hopping strategy based on the array graph configuration set, defining a benefit set based on a common vulnerability scoring system, constructing a honeycomb equation based on the repeated Bayesian game model, and finally optimizing the array graph hopping strategy based on a deep Q network algorithm.

[0034] In some embodiments, in step S1, the technical components can be a container runtime, a cloud native network, a database management system, a container image library, and an orchestration and scheduling system; the attacker action set can be an attack behavior against the container runtime, an attack behavior against the container image library, an attack behavior against the network protocol stack, an attack behavior against the database management system, and an attack behavior against the orchestration and scheduling system, that is, any behavior that attempts to compromise the confidentiality, integrity or availability of these technical components can be regarded as an attack behavior. In addition, the national vulnerability database is a comprehensive vulnerability knowledge base that records in detail the known vulnerabilities and corresponding common vulnerability disclosures of various technical components in the cloud native technology stack. These records cover a wide range of information from vulnerability details to attack techniques that attackers can exploit, providing potential system penetration paths for attackers, so that the common vulnerability disclosure is used to simulate the attack actions that attackers may take in the cloud native environment.

[0035] In some embodiments, in step S1, the array graph configuration set is constructed by three types of technical components, namely a container runtime, a cloud native network and a database management system, and the corresponding technical component sets are , , At least one component is independently selected from each set of technical components to form an effective matrix configuration; therefore, an effective matrix configuration can be... , ... These effective array configurations constitute the array configuration set. Number of effective array configurations ,when hour, The maximum value is 27.

[0036] In some embodiments, in step S1, the Bayesian game model is repeated to define defense and attack strategies, which are respectively defense sets. and attack set ( Among them, the repeated Bayesian game model is defined. It is a defensive collection A finite set of array patterns, representing a transition from the current array pattern configuration to the next valid array pattern configuration, such as... This represents the defender's action of choosing to switch to another effective grid configuration in the current state; repeating the definition of the Bayesian game model. Representing different attacker types A limited set of actions, in which specific actions are disclosed using common vulnerabilities present in the technical components; furthermore... Indicates attacker The probability of launching an attack ( ),in Indicates the first Attacker types, attacker types Represented by a triple: The first element represents the attacker type name, the second element is a two-dimensional value representing the attacker's technical expertise, ranging from [0, 10], and the third element is the attacker's... The probability of launching an attack The attacker type set has a corresponding discrete probability distribution. Since these attacker types and their associated discrete probability distributions are specific to a certain technology component, they can be defined by cloud-native domain security experts and used as input to a text model.

[0037] In some embodiments, in step S1, for the attacker's action set, as long as the attack action... It affected the array configuration. If at least one type of technical component is present, then... Specifically, the function for determining whether an attack is successful is defined based on the following two points. : Firstly, in a specific offensive and defensive phase, the attack action... Impacted technical components Whether the array configuration at this time ( ) at least one technical component used Matching; second, the type of attacker For the professional ability of the impact technology Contrast with the exploitability score of the corresponding common vulnerability and exposure ( ).

[0038] In some embodiments, in step S2, the array graph jump strategy is the selection of the next effective array graph configuration by the honeypot array in the given current array graph configuration The decision process of the next effective array graph configuration (where ). For example, if represents the probability of selecting under the given deployment state The jump strategy is ∑ c ∈ C p c =1 ∀ p c ∈ [0,1] .

[0039] In some embodiments, in step S3, when optimizing the array graph jump strategy based on the deep Q network algorithm, the Bayesian game model of the attack and defense sides is repeated, and the impact score (IS), the exploitability score (ES), and the base score (BS) of the vulnerability defined in the latest vulnerability scoring system v3.1 are used to quantify the benefits of the game. Wherein the impact score IS represents the impact of the attacker taking a specific attack on the confidentiality , integrity and availability of the system, the calculation formula is IS = 6.42 * [1 - (1 - C) * (1 - I) * (1 - A)] ; the exploitability score ES represents the complexity of the attack vector , attack complexity , permission requirement degree and user interaction degree of the four aspects when exploiting a specific vulnerability, the specific calculation formula is: ; the base score BS is composed of the impact score IS and the exploitability score ES. Finally, after setting up the corresponding attack set for each type of attacker, based on the vulnerability score of the vulnerability scoring system, the benefit set of the game of both sides can be defined as follows:

[0040] Attack benefits​ There are two cases for whether the attack is successful: one is when the attack is successful, the attack reward is represented by the base score of the vulnerability scoring system ; the other is when the attack is not successful, the attack reward is represented by the exploitability score of the vulnerability scoring system , which represents the time and cost required for the attacker to perform the attack, and the attack reward is negative at this time. Therefore, the attack reward can be represented as:

[0041] ,

[0042] wherein, represents the attack reward obtained by performing the attack action under the array configuration ; and is used to determine whether the attack is successful, when , it represents that the attack is successful, and when , it represents that the attack is not successful.

[0043] For the defense reward , there are two cases for whether the defense is successful: one is when the defense is not successful, the defense reward is represented by the impact score of the vulnerability scoring system , which represents the impact on the system after the vulnerability is successfully exploited, and the defense reward is negative at this time; the other is when the defense is successful, the defense reward is set to zero, which means that the honeynet will not obtain a positive reward because of the successful defense. Therefore, the defense reward can be represented as:

[0044] ,

[0045] wherein, represents the defense reward obtained by performing the attack action under the array configuration ; and is used to determine whether the defense is successful, when , it represents that the defense is not successful, and when , it represents that the defense is successful.

[0046] In addition, when the attacker does not take any action (i.e., no operation), the reward is zero.

[0047] In some embodiments, when the array hopping strategy is optimized based on the deep Q network algorithm in step S3, a honeynet target equation is constructed based on the reward set and the repeated Bayesian game model to maximize the expected utility of the honeynet, and the equation is as follows:

[0048] ,

[0049] wherein, Honey Array believes that different types of attackers The prior probability, Indicates attacker Select attack action The probability, This represents the hybrid strategy of honeycomb arrays. For the offensive and defensive benefits of the honeycomb formation, This is the cost coefficient. This indicates that the honeycomb array has changed its configuration from the current array diagram. Switch to another array jump configuration The cost of switching at that time Indicates a change in configuration from the current array diagram. Switch to another array jump configuration The corresponding probability.

[0050] Among them, the cost of changing the array configuration can be used Matrix representation, where Row (and column) representation System configuration, cell This indicates that the honeycomb array is configured from the current array diagram. Switch to another array configuration The cost of switching at that time The range is [0, 10], configuration and The greater the heterogeneity between them, the higher the cost of switching.

[0051] In some embodiments, the Deep Q-Network algorithm in step S3 is a deep reinforcement learning algorithm that addresses the dimensionality explosion problem in traditional Q-learning algorithms by combining experience replay and a fixed Q-value objective. The Deep Q-Network algorithm uses two key components: an evaluation network and a target network. The evaluation network is responsible for generating Q-value predictions for all possible actions in the current state, while the target network is used to stabilize the training process by generating a target Q-value to assist in updating the parameters of the evaluation network. In its implementation, the algorithm first initializes the attack / defense environment and network parameters. Then, in each training cycle, the agent interacts with the environment to collect training samples. These training samples are then used to update the network parameters, and the loss function is optimized using gradient descent. The algorithm iterates continuously until a predetermined number of training steps is reached or a convergence condition is met, specifically including the following steps:

[0052] Step 1: Initialize the attack and defense environment model parameters and the experience replay buffer. ( ), evaluate network parameters and target network parameters The parameters of the attack and defense environment model include the array configuration set, attack strategy, defense strategy, attacker's attack probability and payoff set;

[0053] Step 2, initialize the environment and state in each training cycle and according to the prior probability extract the attacker and randomly extract the attack action set of this type of attacker;

[0054] Step 3, select the graph action based on the ε-greedy strategy where the graph action is randomly selected with a probability of ε , otherwise select where represents the action with the maximum Q value, represents the Q value calculated using the evaluation network parameters in state and action ;

[0055] Step 4, calculate the corresponding reward based on the repeated Bayesian game model and the reward set and store the training data ( ) in the experience replay buffer , where represents the current state, represents the next state;

[0056] Step 5, randomly sample training samples from the experience replay buffer , and calculate the Q value of the current state based on the evaluation network and the target Q value of the next state based on the target network, using the following formula:

[0057] Q D = ∑ t=1 ∞ [ γ t - 1 r D ( c t , d t , a t )| c 1 = c; θ] ,

[0058] ,

[0059] where, represents the cumulative defense reward, represents the discount factor, represents the immediate defense reward, represents the initial state , represents the evaluation network parameters; represents the target Q value, represents an immediate reward, represents a discount factor, represents using target network parameters in the next state , defensive action and attack action target Q value, represents target network parameters, represents the number of training samples extracted from the experience replay buffer each time;

[0060] Step 6, calculate the loss function and use gradient descent to backpropagate the loss to update the evaluation network parameters , when the evaluation network parameters the number of update iterations reaches times, the evaluation network parameters are assigned to the target network parameters , wherein represents the interval step number of target network parameter synchronization, and the loss function used is as follows:

[0061] L(θ) = E[ y i - Q(c, d i , a i ; θ) ) 2 ] ,

[0062] wherein, represents the mean square loss of the time difference error, represents the target Q value, represents the evaluation network Q value calculated using the evaluation network parameters in the state , defensive action and attack action , represents the target Q value, represents the number of training samples extracted from the experience replay buffer each time;

[0063] Repeat steps 2 to 6 until the graph hopping strategy converges, output the optimal graph hopping strategy , wherein the number of experience steps required for the algorithm to converge is 100k.

[0064] Referring to Figure 2 , the present application provides a graph hopping system for cloud native honeycomb, comprising the following steps:

[0065] The model construction module 100 acquires an attacker action set and constructs a graph configuration set according to technical components in a cloud native technology stack, and respectively constructs an attack set and a defense set based on a repeated Bayesian game model;

[0066] The strategy generation module 200 generates a graph hopping strategy based on the attack set and the defense set;

[0067] The strategy optimization module 300 optimizes the graph hopping strategy based on a deep Q network algorithm.

[0068] Although the embodiments of the present application have been described in detail above, it is obvious to those skilled in the art that various modifications and changes can be made to the embodiments. However, it should be understood that such modifications and changes are within the scope and spirit of the present application as described in the claims. Moreover, the present application described herein can have other embodiments and can be implemented or realized in various ways.

Claims

1. A method for changing the array diagram of a cloud-native honeycomb array, characterized in that, include: Based on the common vulnerability disclosure data recorded in the national vulnerability database, this study simulates the attack actions an attacker would take in a cloud-native environment to obtain an attacker action set. It also constructs a matrix configuration set based on the technical components in the cloud-native technology stack and builds attack sets based on a repeated Bayesian game model. and defense collection ; When constructing a matrix configuration set based on the technical components in the cloud-native technology stack, it includes: independently selecting at least one component from each of the three types of technical components—container runtime, cloud-native network, and database management system—to form different matrix configurations, and the different matrix configurations ultimately constitute the matrix configuration set; Among them, the definition of the repeated Bayesian game model It is a defensive collection A finite set of matrix diagrams, and the definition Representing different attacker types A finite set of actions; the attacker type Represented by a triple as The first element represents the attacker type name, the second element is a two-dimensional value representing the attacker's technical expertise, and the third element represents the attacker's... The probability of launching an attack ; Based on the attack set and the defense set Generate a pattern transition strategy; Optimizing the graph transition strategy based on the deep Q-network algorithm includes: quantifying the payoffs of the attack set and the defense set using a general scoring system to generate a payoff set, which includes attack payoffs and defense payoffs; constructing a honeycomb objective equation based on the payoff set and a repeated Bayesian game model; and solving for the optimal graph transition strategy using the deep Q-network algorithm. The attack payoffs are as follows: ， in, Indicates the gains from the attack. Indicates the configuration in the array diagram Execute attack actions The gains from the attack; Used to determine whether an attack was successful, denoted as , Indicates an attack action Affected technical components This refers to the technical components in the cloud-native technology stack. Indicates attacker type The impact on technical expertise, when When, it means the attack was successful. When the time is up, it means the attack was unsuccessful; and These represent the base score and the exploitability score of the vulnerability scoring system, respectively. The benefits of the defense are as follows: ， in, Indicates defensive benefits; Indicates the configuration in the array diagram Execute attack actions The defensive benefits gained; Used to determine whether the defense was successful, represented as , Indicates an attack action Affected technical components This refers to the technical components in the cloud-native technology stack. Indicates attacker type The impact on technical expertise, when When, it means the defense was unsuccessful. When this occurs, it signifies a successful defense; This indicates the impact score of the vulnerability scoring system; The target equation for the honey array is as follows: , in, Honey Array believes that different types of attackers The prior probability, Indicates attacker Select attack action The probability, This represents the hybrid strategy of honeycomb arrays. For the offensive and defensive benefits of the honeycomb formation, This is the cost coefficient. This indicates that the honeycomb array has changed its configuration from the current array diagram. Switch to another array jump configuration The cost of jumping at that time Indicates a change in configuration from the current array diagram. Switch to another array jump configuration The corresponding probability.

2. The array pattern transition method according to claim 1, characterized in that, The technical components include container runtime, cloud-native network, database management system, container image library, and orchestration and scheduling system.

3. The array pattern transition method according to claim 2, characterized in that, The attacker's action set includes attacks targeting container runtimes, container image repositories, network protocol stacks, database management systems, and orchestration and scheduling systems.

4. The array pattern transition method according to claim 1, characterized in that, The array pattern transition strategy is a decision-making process that dynamically adjusts the current array pattern configuration to the next effective array pattern configuration based on probability distribution.

5. The array pattern transition method according to claim 1, characterized in that, When solving the optimal matrix transition strategy based on the deep Q-network algorithm, the specific solution process is as follows: Step 1: Initialize the attack and defense environment model parameters, experience replay buffer, evaluation network parameters and target network parameters. The attack and defense environment model parameters include the array configuration set, attack strategy, defense strategy, attacker's attack probability and payoff set. Step 2: In each training cycle, initialize the environment and state, and extract the attacker and their attack action set according to the prior probability; Step 3: Select the action of the array based on the ε-greedy policy, wherein the action of the array is randomly selected with probability ε, otherwise the action with the best Q value in the current evaluation network is selected. Step 4: Calculate the corresponding reward based on the repeated Bayesian game model and the payout set, and store the training data in the experience replay buffer; Step 5: Randomly sample training samples from the experience replay buffer, and calculate the Q value of the current state based on the evaluation network and the target Q value of the next state based on the target network; Step 6: Calculate the loss function and update the parameters of the evaluation network; Repeat steps 2 to 6 until the diagram transition strategy converges, and output the optimal diagram transition strategy.

6. A pattern switching system for cloud-native honeycomb arrays, characterized in that, include: The model building module, based on common vulnerability disclosure data recorded in the national vulnerability database, simulates the attack actions an attacker would take in a cloud-native environment to obtain an attacker action set, and constructs a matrix configuration set based on the technical components in the cloud-native technology stack, and builds attack sets based on a repeated Bayesian game model. and defense collection ; When constructing a matrix configuration set based on the technical components in the cloud-native technology stack, it includes: independently selecting at least one component from each of the three types of technical components—container runtime, cloud-native network, and database management system—to form different matrix configurations, and the different matrix configurations ultimately constitute the matrix configuration set; Among them, the definition of the repeated Bayesian game model It is a defensive collection A finite set of matrix diagrams, and the definition Representing different attacker types A finite set of actions; the attacker type Represented by a triple as The first element represents the attacker type name, the second element is a two-dimensional value representing the attacker's technical expertise, and the third element represents the attacker's... The probability of launching an attack ; The strategy generation module, based on the attack set and the defense set Generate a pattern transition strategy; The strategy optimization module optimizes the diagram transition strategy based on a deep Q-network algorithm, including: quantifying the payoffs of the attack set and the defense set using a general scoring system to generate a payoff set, which includes attack payoffs and defense payoffs; constructing a honeycomb objective equation based on the payoff set and a repeated Bayesian game model; and solving for the optimal diagram transition strategy using the deep Q-network algorithm; the attack payoffs are as follows: ， in, Indicates the gains from the attack. Indicates the configuration in the array diagram Execute attack actions The gains from the attack; Used to determine whether an attack was successful, when When, it means the attack was successful. When the time is up, it means the attack was unsuccessful; and These represent the base score and the exploitability score of the vulnerability scoring system, respectively. The benefits of the defense are as follows: ， in, Indicates defensive benefits; Indicates the configuration in the array diagram Execute attack actions The defensive benefits gained; Used to determine whether the defense was successful, when When, it means the defense was unsuccessful. When this occurs, it signifies a successful defense; This indicates the impact score of the vulnerability scoring system; The target equation for the honey array is as follows: , in, Honey Array believes that different types of attackers The prior probability, Indicates attacker Select attack action The probability, This represents the hybrid strategy of honeycomb arrays. For the offensive and defensive benefits of the honeycomb formation, This is the cost coefficient. This indicates that the honeycomb array has changed its configuration from the current array diagram. Switch to another array jump configuration The cost of jumping at that time Indicates a change in configuration from the current array diagram. Switch to another array jump configuration The corresponding probability.

Images