A method for scheduling executors based on confidence and heterogeneity

Abstract

This invention proposes a method for scheduling executables based on confidence and heterogeneity. The confidence and heterogeneity of the executable are initialized using known vulnerability information of the executable's internal components. When inconsistent executable output occurs, the executable is first tested using pre-set test samples. If an executable error occurs, it can be determined that the executable has been attacked, and based on the test results, a first scheduling solution is adopted. Otherwise, it is uncertain whether an executable has been attacked, and a second scheduling solution is adopted based on the confidence and heterogeneity of the executable, thereby ensuring that the system's external characteristics remain dynamic and uncertain. After scheduling is completed, the confidence and heterogeneity of the executable are adjusted.

Description

Technical Field

[0001] The present invention relates to an execution body scheduling method based on confidence and heterogeneity, and belongs to the technical field of mimicry defense. Background Art

[0002] The core of mimetic defense technology is a dynamic, heterogeneous, redundant architecture, characterized by three key characteristics: dynamism, heterogeneity, and redundancy. Dynamicity means that when a system is attacked by a cyberattack, the scheduling module will schedule executors according to pre-set security policies, preventing attackers from penetrating the system's internal defense architecture. Heterogeneity means that functionally equivalent executors may differ, for example, based on underlying hardware, operating systems, protocols, or programming languages. Redundancy means that multiple executors can run simultaneously online, with a pool of redundant backup executors.

[0003] Mimicry defense technology primarily utilizes the executor scheduling module to achieve the dynamic and uncertain nature of the system's external characteristics. Based on pre-set security policies, the scheduler removes abnormal executors from the running state and selects suitable backup executors from the executor pool to bring them online, thereby changing the system's current operating state and rendering the attack ineffective. Therefore, executor scheduling strategies play a crucial role in mimicry defense technology. Designing an optimal executor scheduling strategy that addresses executor heterogeneity while also ensuring executor reliability is a worthy topic of discussion. Summary of the Invention

[0004] To address the aforementioned issue of executor scheduling strategies, this paper proposes an executor scheduling method based on confidence and heterogeneity. The confidence and heterogeneity of the executor are initialized using known vulnerability information from its internal components. When inconsistent executor output occurs, the executor is first tested using pre-set test cases. Then, based on the test results and the executor's confidence and heterogeneity, the executor is scheduled in a targeted manner to ensure that the system's external characteristics remain dynamic and uncertain. Finally, the confidence and heterogeneity of the executor are adjusted.

[0005] In order to achieve the above invention purpose, the present invention adopts the following technical solutions:

[0006] A method for scheduling an executable based on confidence and heterogeneity, the method comprising the following steps:

[0007] Step 1: First, based on the known vulnerability information of the components within the executable, the confidence of all components is calculated to achieve component reliability measurement. Then, based on the component confidence, the initial confidence of the executable is calculated, and the evaluation granularity of the executable confidence is refined. Finally, the heterogeneity between the executables is calculated based on the component type and vulnerability information.

[0008] Step 2: When inconsistent outputs from executables occur, the executables are first tested in real time using pre-set test samples. If an executable has an error, the first scheduling solution is adopted based on the test results. Otherwise, the second scheduling solution is adopted based on the confidence and heterogeneity of the executable.

[0009] Step 3: Dynamically adjust the confidence and heterogeneity of the executor based on the output of the executor and the test results.

[0010] Furthermore, the calculation method of the component confidence and the setting method of the initial confidence of the execution body in step 1 are as follows:

[0011] (1.1) Assume that an executable consists of L layers of components, where the first layer has M different components, denoted by C = {c1, c2, ..., c M};

[0012] (1.2) Assume that all components in set C contain H different types of vulnerabilities, denoted as Vul = {vul1, vul2,…, vul H}, component c i Contains h i vulnerabilities, denoted as

[0013] (1.3) The Common Vulnerability Scoring System (CVSS) is used to quantify the threat level of vulnerabilities. The quantitative score range is [0, 10]. The higher the score, the higher the threat level of the vulnerability. The "total threat value of all vulnerabilities at layer l" is defined as:

[0014]

[0015] Define "component c i The threat value of the contained vulnerabilities is:

[0016]

[0017] (1.4) Define component c i The original confidence is the sum of the threat value of all vulnerabilities in layer l and the value of component c i The ratio of the threat value of the contained vulnerabilities is:

[0018]

[0019] Where τ∈[0,10] is the adjustment factor. The smaller τ is, the greater the difference in initial confidence between different executors; the larger τ is, the smaller the difference in initial confidence between different executors.

[0020] (1.5) Normalize the original confidence of the component to be the initial confidence of the component:

[0021]

[0022] (1.6) Assume that the execution body X={x1,x2,…,x L It consists of L layers of components, and defines the initial confidence of the executive body X as:

[0023]

[0024] where α l is the component weight coefficient, which indicates the importance of each component in the entire execution body.

[0025] Furthermore, the method for setting the initial heterogeneity of the executable in step 1 is as follows:

[0026] (1.7) If two components are of different types (e.g., Windows and Linux), then the heterogeneity between them is directly set to 1. If two components are of the same type (e.g., Win7 and Win10), then the heterogeneity is calculated based on the vulnerability information between them, as follows:

[0027] (1.8) Assume that component c i Contains v i vulnerabilities, component c j Contains v j vulnerabilities, and they all contain v ij vulnerabilities, defining component c i With component c j The degree of heterogeneity between them is:

[0028]

[0029] (1.9) Assume that the execution body X={x1,x2,…,x L} and Y={y1,y2,…,y L} consists of L-layer components, and its initial heterogeneity is defined as:

[0030]

[0031] where α l is the component weight coefficient, which indicates the importance of each component in the entire execution body.

[0032] Furthermore, the execution body scheduling method in step 2 is as follows:

[0033] (2.1) Use test cases with known correct answers to test the online executable.

[0034] (2.2) If an error occurs in an executable, it can be determined that the executable has been attacked, and the process jumps to (2.3) to adopt the first scheduling plan. Otherwise, it is not certain whether any executable has been attacked, so the process jumps to (2.7) to adopt the second scheduling plan.

[0035] The first scheduling scheme of the executor is as follows:

[0036] (2.3) Let the set of “executables with errors” be E;

[0037] (2.4) Let the set of “outputs of the executables in set E” be F;

[0038] (2.5) For an executable outside the set E, if its output belongs to the set F, then add the executable to the set E;

[0039] (2.6) All executors in the set E are taken offline for cleaning. At the same time, a backup executor is selected from the executor pool based on the confidence level and heterogeneity level to be put online. The specific selection method is shown in (2.11);

[0040] The second scheduling scheme of the executor is as follows:

[0041] (2.7) Divide the set according to the answers given by the executors, and calculate the heterogeneity evaluation index between each pair of executors in each set:

[0042] Het(X,Y),stX,Y∈S k ,X≠Y,k=1,2,…,K

[0043] Where K is the number of sets, that is, the number of types of answers given by the executor;

[0044] (2.8) Calculate the score of each set using the following formula:

[0045]

[0046] The meaning of this formula is that, based on the confidence of the executor, if its heterogeneity with other executors in the set exceeds 0.5, then its contribution to the score is increased, otherwise its contribution to the score is decreased;

[0047] (2.9) The set of executors with the highest scores is retained, and all other executors are taken offline for cleaning. At the same time, backup executors are selected from the executor pool based on confidence and heterogeneity to be put online. The specific selection method is shown in (2.10);

[0048] The method for selecting a backup executive is as follows:

[0049] (2.10) Assume that there are N spare executors {X1, X2, ..., X N}, the remaining M execution bodies on the line {Y1,Y2,…,Y M First, calculate the heterogeneity index between each standby executor in the executor pool and the remaining online executors:

[0050]

[0051] (2.11) Matrix Het Find the minimum value along a row:

[0052]

[0053] Min_Het=[δ1 δ2 … δ N ] T ;

[0054] (2.12) Define the heterogeneity threshold ε, if the backup execution body X i The minimum heterogeneity δ i >ε, then add the executable to the candidate executable set T;

[0055] (2.13) Compute the set T = {X1, X2,…, X |T| For each candidate executor in}, the mean heterogeneity of the remaining executors on the line is:

[0056]

[0057] (2.14) Compute the set T = {X1, X2, ..., X |T| The score of each execution body in}:

[0058] Score(X i )=β C ·Conf(X i )+β H ·MeanHet(X i ),i=1,2,…,|T|

[0059] where β C ,β H are all weight coefficients and satisfy β C +β H =1;

[0060] (2.15) First, add the executable with the highest score to the set G, and then sort the remaining candidate executables in the set T from highest to lowest score;

[0061] (2.16) Traverse each execution body of the set T in order. If the execution body X i If the degree of heterogeneity with every executor in the set G is greater than the threshold ε, the executor is added to the set G, otherwise it is discarded;

[0062] (2.17) When the number of executables in set G meets the requirement, or when all the executables in set T are traversed, the executables in set G are put online and the scheduling is completed.

[0063] Furthermore, the adjustment scheme for the confidence and heterogeneity of the execution body in step 3 is as follows:

[0064] Assume that there are K executors running online, KM of them are taken offline for cleaning, and M are retained;

[0065] For the first scheduling scheme of the executor, the adjustment scheme of confidence and heterogeneity is as follows:

[0066] (3.1) Calculate the confidence increase ratio:

[0067]

[0068] Among them, upMax is the upper limit of the increase ratio;

[0069] (3.2) The confidence of the M retained execution bodies is updated as follows:

[0070] Conf(X)=min(confMax,(1+upRatio)·Conf(X))

[0071] Where confMax is the upper confidence limit;

[0072] (3.3) Calculate the confidence reduction ratio:

[0073]

[0074] Among them, downMax is the upper limit of the downward adjustment ratio;

[0075] (3.4) The confidence of KM offline cleaning executors is updated as:

[0076] Conf(X)=max(confMin,(1-downRatio)·Conf(X))

[0077] Where confMin is the confidence lower limit. When the lower limit is reached, the execution body is abandoned.

[0078] (3.5) For the KM offline cleaned executors, divide them into different sets {S1, S2, ...} according to their output categories, and reduce the heterogeneity between the executors in each set by a ratio of where |S i | is the number of elements in the set;

[0079] (3.6) The heterogeneity between the executors in each set is updated as follows:

[0080]

[0081] (3.7) For the M retained execution bodies, keep their heterogeneity unchanged:

[0082] For the second scheduling scheme of the executor, the adjustment scheme of confidence and heterogeneity is as follows:

[0083] (3.8) In the second scheduling scenario, since it is impossible to determine whether any executor is currently under attack, and it is also impossible to determine what the correct output is for the system input, the confidence of the executor is not updated, and only the heterogeneity is updated;

[0084] (3.9) All executables are divided into different sets {S1, S2, ...} according to their output categories, and the heterogeneity between the executables in each set is reduced by a ratio of where |S i | is the number of elements in the set;

[0085] (3.10) The heterogeneity between the execution bodies in each set is updated as:

[0086]

[0087] Compared with the existing technology, the execution body scheduling method based on confidence and heterogeneity described in the present invention has the following advantages:

[0088] (1) By using the known vulnerability information of the internal components of the executive body, the confidence of the components and the heterogeneity between the components are calculated, thus achieving the reliability measurement of the components. On this basis, the confidence and heterogeneity of the executive body are dynamically initialized according to the importance of the components at each layer, and the initialization method is more reasonable;

[0089] (2) When inconsistent outputs of the executable occur, the executable is first tested in real time using pre-set test samples. Then, based on the test results and the confidence and heterogeneity of the executable, the executable is scheduled in a targeted manner to better maintain the dynamic and uncertain characteristics of the system's external features.

[0090] (3) Based on the output of the executor and the test results, the confidence and heterogeneity of the executor are dynamically adjusted to adapt to the ever-changing network environment, thereby improving the overall defense capability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] In order to more clearly illustrate the technical solutions in the present disclosure or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0092] Figure 1 This is a flow chart of an execution body scheduling method based on confidence and heterogeneity disclosed in an embodiment of the present invention;

[0093] Figure 2 A schematic diagram of an executor scheduling method based on confidence and heterogeneity disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0094] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0095] This invention proposes a method for scheduling executables based on confidence and heterogeneity. The confidence and heterogeneity of the executables are initialized using known vulnerability information of the executables' internal components. When inconsistent executable outputs occur, the executables are first tested using pre-set test samples. If an executable gives an incorrect answer, it can be determined that the executable has been attacked, and based on the test results, a first scheduling solution is adopted. Otherwise, it is uncertain whether any executable has been attacked, and a second scheduling solution is adopted based on the executables' confidence and heterogeneity, thereby ensuring that the system's external characteristics remain dynamic and uncertain. After scheduling is completed, the confidence and heterogeneity of the executables are adjusted.

[0096] See also Figure 1 , a flow chart of an execution body scheduling method based on confidence and heterogeneity disclosed in an embodiment of the present invention, the method includes:

[0097] Step S101, calculating component confidence:

[0098] In this embodiment, an executive body is composed of L layers of components, wherein the first layer has M different components, which are denoted as C = {c1, c2, ..., c M};

[0099] All components in set C (i.e., layer l) contain a total of H different types of vulnerabilities, denoted as Vul = {vul1, vul2, ..., vul H}, component ci Contains h i vulnerabilities, denoted as

[0100] The Common Vulnerability Scoring System (CVSS) is used to quantify the threat level of vulnerabilities. The quantitative score range is [0,10]. The higher the score, the higher the threat level of the vulnerability.

[0101] The total threat value of all vulnerabilities at layer l is defined as:

[0102]

[0103] Define "component c i The threat value of the contained vulnerabilities is:

[0104]

[0105] Define component c i The original confidence is the sum of the threat value of all vulnerabilities in layer l and the value of component c i The ratio of the threat value of the contained vulnerabilities is:

[0106]

[0107] thr(c i ) is smaller, indicating that the component c i The smaller the risk level of the component is relative to other components in the same layer, the smaller the original confidence conf′(c i ) should be larger;

[0108] τ∈[0,10] is the adjustment factor. The smaller τ is, the greater the difference in initial confidence between different executors; the larger τ is, the smaller the difference in initial confidence between different executors.

[0109] Normalize the original confidence of the component as the initial confidence of the component:

[0110]

[0111] Step S102, calculating component heterogeneity:

[0112] In this embodiment, if two components are of different types (for example, Windows and Linux), their heterogeneity is directly set to 1; if two components are of the same type (for example, Win7 and Win10), the heterogeneity is calculated based on the vulnerability information between them, as follows:

[0113] Assume component c i Contains v i vulnerabilities, component cj Contains v j vulnerabilities, and they all contain v ij vulnerabilities, defining component c i With component c j The degree of heterogeneity between them is:

[0114]

[0115] When component c i With component c j When the vulnerability information of v is exactly the same, that is, v i =v j =v ij , their heterogeneity het(c i ,c j )=0; when component c i With component c j When the vulnerability information of v is completely different, that is, v ij = 0, their heterogeneity het(c i ,c j )=1.

[0116] Step S103: Calculate the initial confidence and heterogeneity of the execution body:

[0117] In this embodiment, the execution body X={x1, x2, ..., x L} is composed of L layer components;

[0118] Define the initial confidence of the executive body X as:

[0119]

[0120] Define the initial heterogeneity of the execution bodies X and Y as:

[0121]

[0122] where α l is the component weight coefficient, which indicates the importance of each component in the entire execution body.

[0123] Step S104: Use the test sample to test the executable:

[0124] In this example, there are five executables {X1, X2, X3, X4, X5} running online. For the current system input, the output of the executable is {a, b, a, c, b}. Inconsistent outputs occur, requiring real-time detection of the executables.

[0125] Randomly extract T test samples from the sample library and input them into the online execution body for testing.

[0126] If an error occurs in an executable body, it can be determined that the executable body has been attacked, and the process jumps to step S105 to adopt the first scheduling solution; otherwise, it cannot be determined whether any executable body has been attacked, and the process jumps to step S106 to adopt the second scheduling solution.

[0127] Step S105: The first scheduling scheme of the execution body is as follows:

[0128] In this embodiment, errors occurred in the execution bodies X2 and X4 during the test process, which are recorded as the set E = {X2, X4};

[0129] For the system input, the output of the executable in the set E = {X2, X4} is recorded as the set F = {b, c};

[0130] For an executable outside of set E, if its output belongs to set F, the executable is added to set E. In this embodiment, since the output b of executable X5 belongs to set F, it is added to set E = {X2, X4, X5}.

[0131] All executors in the set E are taken offline for cleaning. At the same time, a backup executor is selected from the executor pool based on the confidence and heterogeneity to be put online. The specific selection method is shown in step S107.

[0132] Step S106: The second scheduling scheme of the execution body is as follows:

[0133] Divide the set according to the output of the executable, S1 = {X1, X3}, S2 = {X2, X5}, S3 = {X4};

[0134] Calculate the score of each set using the following formula:

[0135]

[0136] The meaning of this formula is that, based on the confidence of the executor, if its heterogeneity with other executors in the set exceeds 0.5, then its contribution to the score is increased, otherwise its contribution to the score is decreased;

[0137] There is only one executor X4 in the set S3, so the score of the set is equal to the confidence of the executor X4, Score(S3) = Conf(X4);

[0138] In this embodiment, set S1 has the highest score, and the executors {X1, X3} in this set are retained. All other executors are taken offline for cleaning. At the same time, a backup executor is selected from the executor pool based on the confidence and heterogeneity to be put online. The specific selection method is shown in step S107.

[0139] Step S107: The method for selecting a backup execution body is as follows:

[0140] In this example, there are five standby executors {R1, R2, …, R5} in the executor pool, and two online executors {X1, X3}. First, the heterogeneity index between each standby executor in the executor pool and the remaining online executors is calculated:

[0141]

[0142] Matrix Het Find the minimum value along a row:

[0143]

[0144] Min_Het = [δ1 δ2 … δ5] T ;

[0145] Define the heterogeneity threshold ε, if the backup execution body X i The minimum heterogeneity δ i >ε, then add the executable to the candidate executable set T;

[0146] In this embodiment, the set T = {R1, R2, R3, R5}, and the mean heterogeneity of each executable in the set with respect to the remaining online executables {X1, X3} is calculated:

[0147]

[0148] Calculate the score of each executable in the set T = {R1, R2, R3, R5}:

[0149] Score(R i )=β C ·Conf(R i )+β H MeanHet(R i ),i=1,2,3,5

[0150] where β C ,β H are all weight coefficients and satisfy β C +β H =1;

[0151] First, add the executor with the highest score to set G, and then sort the remaining candidate executors in set T from highest to lowest score;

[0152] Traverse each execution body of the set T in order. If the execution body X i If the degree of heterogeneity with every executor in the set G is greater than the threshold ε, the executor is added to the set G, otherwise it is discarded;

[0153] When the number of executables in set G meets the requirement, or when all the executables in set T are traversed, the executables in set G are put online and the scheduling is completed.

[0154] In step S108, the adjustment scheme of the execution volume confidence and heterogeneity is as follows:

[0155] In this embodiment, there are 5 executors {X1, X2, X3, X4, X5} running online, 3 {X2, X4, X5} are cleaned offline, and 2 {X1, X3} are retained;

[0156] For the first scheduling scheme of the executor, the adjustment scheme of confidence and heterogeneity is as follows:

[0157] Calculate the confidence increase ratio:

[0158]

[0159] Among them, upMax is the upper limit of the increase ratio;

[0160] The confidence of the retained execution body is updated as follows:

[0161] Conf(X i )=min(confMax,(1+upRatio)·Conf(X i )),i=1,3

[0162] Where confMax is the upper confidence limit;

[0163] Calculate the confidence reduction ratio:

[0164]

[0165] Among them, downMax is the upper limit of the downward adjustment ratio;

[0166] The confidence level of the offline cleaned executable is updated to:

[0167] Conf(X i )=max(confMin,(1-downRatio)·Conf(X i )),i=2,4,5

[0168] Where confMin is the confidence lower limit. When the lower limit is reached, the execution body is abandoned.

[0169] For the offline cleaned executors, they are divided into different sets S1 = {X2, X5}, S2 = {X4} according to their output, and the heterogeneity between the executors in each set is reduced by a ratio of where |S i | is the number of elements in the set;

[0170] The heterogeneity between the executors in each set is updated as follows:

[0171]

[0172] For the retained execution bodies, their heterogeneity is maintained unchanged;

[0173] For the second scheduling scheme of the executor, the adjustment scheme of confidence and heterogeneity is as follows:

[0174] In the second scheduling scheme, since it is impossible to determine whether any executor is currently under attack, and it is also impossible to determine what the correct output is for the system input, the confidence of the executor is not updated, and only the heterogeneity is updated;

[0175] According to the output, all executables are divided into different sets S1 = {X1, X3}, S2 = {X2, X5}, S3 = {X4}, and the heterogeneity between the executables in each set is reduced by a ratio of where |S i | is the number of elements in the set;

[0176] The heterogeneity between the executors in each set is updated as follows:

[0177]

[0178] The above descriptions are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the scope of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for scheduling executors based on confidence and heterogeneity, characterized in that: The following steps are involved: Step 1: Based on known vulnerability information, calculate the initial confidence and initial heterogeneity of the executable, and refine the evaluation granularity of these two indicators; Step 2: When inconsistent outputs of the executables occur, use the pre-set test samples to test the executables in real time, and adopt different scheduling plans based on whether any executables have errors. Step 3: Dynamically adjust the confidence and heterogeneity of the executor based on the output of the executor and the test results; The scheduling method of the execution body in step 2 specifically includes the following steps: Step 2.1: Use test cases with known correct answers to conduct real-time testing on the online executable. Step 2.2: If an error occurs in an executable, it can be determined that the executable has been attacked. Jump to step 2.3 and adopt the first scheduling plan. Otherwise, it is not certain whether any executable has been attacked. Jump to step 2.6 and adopt the second scheduling plan. Step 2.3: Count all the executables that have errors, record them as set E; count the outputs of the executables in set E, record them as set F; Step 2.4: For any executable outside of set E, if its output belongs to set F, add the executable to set E. Step 2.5: Offline and clean the executors in set E, then jump to step 2.8 to select a backup executor to run online; Step 2.6: Divide the set according to the answers given by the executor, and calculate the score of each set based on the confidence and heterogeneity; Step 2.7: Keep the set of executors with the highest scores, remove all other executors from the network, and then go to step 2.8 to select a backup executor to run online. Step 2.8: The executors to be put into operation must first meet the minimum heterogeneity requirement. Then, based on this, the optimal set of executors is selected by comprehensively considering the confidence and heterogeneity. In step 3, the confidence and heterogeneity are dynamically adjusted as follows: For the first scheduling scheme, the confidence of the retained executors is increased, and the confidence and heterogeneity of the executors that are offline for cleaning are decreased. For the second scheduling scheme, since it is impossible to determine whether any executors have been attacked, and it is also impossible to determine what the correct output is for the system input, the confidence of the executors is not updated, and only the heterogeneity between the executors is decreased.

2. The method for scheduling an executable based on confidence and heterogeneity according to claim 1, characterized in that: The calculation of the initial confidence and initial heterogeneity of the execution volume in step 1 specifically includes the following steps: Step 1.1: Quantify the threat level of each vulnerability using the Common Vulnerability Scoring System. Step 1.2: Calculate the confidence of each component and the heterogeneity between components; Step 1.3: Calculate the confidence of each executor and the degree of heterogeneity between executors.

Images