A quantum computing side channel defense method based on scheduler output observability control

Abstract

The application discloses a quantum computing side channel defense method based on scheduler output observability control, obtains visible output data returned by a quantum operation system scheduler to a tenant, constructs an output observation track composed of waiting time, starting time, completion time, task state, backend selection, acceptance / rejection state or retry state, etc.; determines a leakage risk score according to the time sequence granularity of the observation track, the backend distinguishability, the output stability and the detection adaptability; when the leakage risk meets a preset condition, selects and executes time sequence output coarsening, backend selection masking, detection response suppression, flow limiting cooling and access parameter updating strategies, and converts the original visible output into a controlled visible output. The application can reduce the ability of an attacker to infer the hidden state of the scheduler, the backend selection or the existence of a protected job in a multi-tenant quantum cloud environment by repeatedly submitting detection jobs, and improve the scheduling security and stability of quantum computing services.

Description

Technical Field

[0001] This invention relates to the fields of quantum computing security and quantum operating system scheduling technology, and in particular to a quantum computing side-channel defense method based on the observability control of scheduler output. Background Technology

[0002] As quantum computing gradually moves from single-user experimental devices to cloud-based services, quantum operating systems or quantum cloud runtimes need to handle job reception, queue sorting, resource allocation, backend selection, execution control, and result return among quantum computing jobs submitted by multiple tenants. The scheduler thus becomes a key component connecting user jobs with limited quantum backend resources.

[0003] In a multi-tenant quantum cloud environment, different tenants typically do not trust each other, while quantum backends are characterized by scarce resources, significant topological constraints, calibration status changes over time, and execution quality being affected by queuing and co-location. The output information generated by the scheduler when processing jobs, such as waiting time, start time, completion time, rejection or retry status, and backend selection results, while necessary for service interaction, may also reflect the scheduler's internal state, resource pressure, other tenant job types, or the isolation requirements of protected jobs.

[0004] Existing quantum computing services or quantum operating system scheduling schemes typically focus more on throughput, latency, execution fidelity, and resource utilization, with insufficient control over the observability of the scheduler's output. Attackers do not need to directly compromise the quantum backend or scheduler; they can simply submit probe jobs repeatedly through ordinary tenant interfaces and observe job responses, queuing delays, start or completion delays, and backend selection changes. This can create timing side channels or output side channels, allowing attackers to infer hidden scheduling states or job attributes of other tenants.

[0005] Therefore, it is necessary to propose an output observability control method for quantum operating system schedulers, which treats the scheduler's visible output as a manageable and secure object. By using mechanisms such as timing output coarsening, backend selection masking, and probe response suppression, the side-channel information content in the visible output is reduced, while maintaining the availability of quantum cloud services and normal operating experience as much as possible. Summary of the Invention

[0006] The purpose of this invention is to provide a quantum computing side-channel defense method based on the observability control of scheduler output, in order to solve the problem that the side-channel leakage risk is high due to the overly fine scheduler output, exposure of back-end selection information, and lack of suppression of probe job responses during the scheduling process of existing quantum operating systems or quantum clouds.

[0007] To address the above problems, this invention provides a quantum computing side-channel defense method based on scheduler output observability control, comprising:

[0008] Obtain the visible output data of the scheduler of the quantum operating system; determine the probe trajectory of the scheduler based on the visible output data; and determine whether there is a risk of leakage based on the time domain change of the probe trajectory.

[0009] When there is a risk of leakage, the suspicious party of the scheduler is identified based on the job submission history information of the scheduler; the job submission of the suspicious party is traced to determine whether the suspicious party is an attacker;

[0010] Monitor the attacker's interface interaction logs on the quantum operating system, determine the attacker's job submission characteristics based on the interface interaction logs, and determine a defense strategy against the attacker based on the job submission characteristics.

[0011] Based on the aforementioned defense strategy, adjust the job interception operation of the quantum computing side channel.

[0012] Optionally, the visible output data of the quantum operating system's scheduler is acquired, and the probe trajectory for the scheduler is determined based on the visible output data, including:

[0013] Obtain several visible output data of the quantum operating system scheduler for several input tasks, each corresponding to a specific input task; wherein, each visible output data includes the scheduler's waiting time, start time, completion time, and task processing result for each input task.

[0014] The visible output data are transformed to obtain the processing representation vector of each input task; the processing representation vectors of all input tasks are sequentially arranged and integrated to obtain the detection trajectory of the scheduler.

[0015] Optionally, based on the time-domain changes of the detection trajectory, determining whether the detection trajectory poses a risk of leakage includes:

[0016] The temporal changes of all processing representation vectors under the detection trajectory are compared to obtain several element change queues; wherein, the elements include the waiting time, start time, completion time, and task processing result of each input task;

[0017] Determine whether each element change queue meets a preset fine-grained condition; if at least one element change queue meets the preset fine-grained condition, then determine that the detection trajectory has a risk of information leakage; otherwise, determine that the detection trajectory does not have a risk of information leakage.

[0018] Optionally, when there is a risk of leakage, based on the scheduler's job submission history information, suspicious individuals associated with the scheduler can be identified, including:

[0019] If there is a risk of leakage, the job submission log is obtained according to the job submission interface currently connected to the scheduler;

[0020] The job submission logs are filtered by historical time intervals to obtain job submission history information; the job submission history information is then analyzed to identify time-domain changes in job submission status, and users who meet the job submission conditions of the scheduler are identified as suspicious users of the scheduler; wherein, the job submission status includes the type of submitted job, time distribution, and user identity.

[0021] Optionally, the job submission of the suspected individual is traced to determine whether the suspected individual is an attacker, including:

[0022] Obtain the job submission behavior information of the suspected individual at the job submission interface; wherein, the job submission behavior information includes the time information of each job submission by the suspected individual;

[0023] Based on the job submission behavior information, determine whether the jobs submitted by the suspicious person in the past have caused service delays for the scheduler; if so, determine that the suspicious person is an attacker; otherwise, determine that the suspicious person is not an attacker.

[0024] Optionally, the attacker's interface interaction logs on the quantum operating system are monitored, and the attacker's job submission characteristics are determined based on the interface interaction logs, including:

[0025] Based on the attacker's identity information, the quantum operating system extracts the attacker's interface interaction logs; wherein, the interface interaction logs include records of the attacker's interaction behavior with the quantum operating system's submission interface;

[0026] Extract the temporal and content characteristics of the attacker's job submission to the submission interface from the interface interaction log, and use these characteristics as job submission features; wherein, the temporal and content characteristics include the time point and the amount of job data of each job submission initiated by the attacker.

[0027] Optionally, based on the job submission characteristics, a defense strategy against the attacker is determined, including:

[0028] Based on the time domain and content characteristics, the occupancy status of the quantum computing side channel of the quantum operating system by the job submitted by the attacker is determined; wherein, the occupancy status includes the time interval of the attacker's occupation of the channel and the type of resources occupied;

[0029] Based on the occupancy status, a defense strategy for the channel against the attacker is determined; wherein the defense strategy includes a strategy for restricting the attacker's use of the channel.

[0030] Optionally, the job interception operation of the quantum computing side channel is adjusted according to the defense strategy, including:

[0031] Based on the channel restriction policy for the attacker included in the defense strategy, the interception operation parameters of the quantum computing side channel for the jobs submitted by the attacker are adjusted; wherein, the interception operation parameters include the duration of the interception operation and the type and quantity of resources that are prohibited from being allocated corresponding to the interception operation.

[0032] Optionally, it further includes adjusting the resource configuration of the channel according to the execution status of the job interception operation, which includes:

[0033] The execution progress of the job interception operation is obtained to determine whether the job interception operation has been completed.

[0034] Once the job interception operation is completed, the resource configuration status of the channel for other jobs is adjusted; wherein, the resource configuration status includes the resource allocation type and quantity for other jobs;

[0035] If the interception operation is not completed, the current interception operation parameters for the channel remain unchanged.

[0036] This invention provides a quantum computing side-channel defense method based on scheduler output observability control. The method acquires visible output data generated by the quantum operating system scheduler during the processing of multiple quantum computing jobs; constructs an output observation trajectory based on the visible output data and job submission information; determines the scheduler output observability leakage risk based on the output observation trajectory; when the leakage risk meets preset defense trigger conditions, determines at least one of a timing output coarsening strategy, a backend selection masking strategy, and a probe response suppression strategy; adjusts the visible output returned by the scheduler to the job submission entity according to the strategy, and updates the scheduler's admission control parameters, limit control parameters, or output granularity control parameters based on the adjustment result.

[0037] Compared with the prior art, the present invention has at least the following beneficial effects:

[0038] First, the visible output of the scheduler is used as the target of side-channel defense to avoid relying solely on post-event attacker identification or simple interception operations.

[0039] Second, it can coarsen or release fine-grained timing information such as waiting time, start time, and completion time in batches, thereby reducing the timing inference capability.

[0040] Third, it can mask specific quantum back-end identifiers and candidate back-end sets, reducing the exposure of back-end selection results to internal resource pressure;

[0041] Fourth, it can suppress the response of high-frequency, periodic, or adaptive detection operations, thereby reducing detection gains;

[0042] Fifth, this method can be extended to multi-tenant quantum cloud environments and can dynamically adjust the output granularity according to service level and tenant-level risk status.

[0043] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0044] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0045] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0046] Figure 1 This is a flowchart illustrating a quantum computing side-channel defense method based on scheduler output observability control provided in an embodiment of the present invention.

[0047] Figure 2 This is a specific embodiment of the quantum computing side-channel defense method based on scheduler output observability control of the present invention.

[0048] Figure 3 This is a schematic diagram of the output observability control structure in a multi-tenant quantum cloud environment provided by an embodiment of the present invention.

[0049] Figure 4 This is a schematic diagram illustrating the operation of an output observability control strategy provided in an embodiment of the present invention. Detailed Implementation

[0050] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0051] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," "far," "near," "front," and "rear," etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0052] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0053] Please see Figure 1 As shown, an embodiment of this application provides a quantum computing side-channel defense method based on scheduler output observability control. This quantum computing side-channel defense method based on scheduler output observability control includes:

[0054] Obtain the visible output data of the quantum operating system's scheduler, determine the probe trajectory of the scheduler based on the visible output data, and judge whether there is a risk of leakage based on the time domain changes of the probe trajectory.

[0055] When there is a risk of leakage, the scheduler's job submission history information is used to identify suspicious individuals on the scheduler; the job submissions of the suspicious individuals are traced back to determine whether they are attackers.

[0056] Monitor the attacker's interface interaction logs on the quantum operating system, determine the attacker's job submission characteristics based on the interface interaction logs, and determine the defense strategy against the attacker based on the job submission characteristics.

[0057] Adjust the job interception operation of the quantum computing side channel according to the defense strategy.

[0058] The beneficial effects of the above embodiments are that the quantum computing side-channel defense method based on the observability control of the scheduler output determines whether there is any information snooping behavior against the scheduler by tracing and analyzing the visible output of the scheduler, thereby identifying the attackers of the scheduler, facilitating subsequent accurate defense against the attackers' snooping behavior, avoiding the leakage of scheduler information and ensuring the security of the scheduler.

[0059] In another embodiment, the visible output data of the quantum operating system's scheduler is obtained, and the probe trajectory of the scheduler is determined based on the visible output data, including:

[0060] Obtain several visible output data of the quantum operating system's scheduler for several input tasks, each corresponding to a specific task; wherein, each visible output data includes the scheduler's waiting time, start time, completion time, and task processing result for each input task;

[0061] The visible output data is transformed to obtain the processing representation vector of each input task; the processing representation vectors of all input tasks are then arranged and integrated in a temporal order to obtain the probe trajectory of the scheduler.

[0062] Attackers typically submit probe jobs repeatedly to the scheduler. Upon receiving these probe jobs, the scheduler processes them according to its own operational status and resource scheduling decisions. Understandably, these probe jobs are treated equally with other normal jobs received by the scheduler at the job processing level. The scheduler does not know in advance whether a received job is a probe job, therefore it does not differentiate between probe jobs and other normal jobs. It allocates computing resources to each job (including probe jobs and other normal jobs) according to its actual operational status and existing resource scheduling decisions. The attacker's purpose in submitting probe jobs is not to actually require the scheduler to process them, but to interfere with and influence the scheduler's operational status. The scheduler's own processing of each job is theoretically hidden and invisible, but the results are output in a visible form. If the attacker submits a sufficient number of probe jobs, the submission time follows a certain periodic pattern, and the data content meets the corresponding data volume and type conditions, the attacker can collect and analyze the scheduler's processing results for all probe jobs to deduce the scheduler's operational status and resource scheduling decisions, leading to the leakage of the scheduler's information.

[0063] To prevent attackers from spying on the scheduler's information using the methods described above, it is necessary to identify the attacker's probe trajectory during the scheduler's operation. The jobs submitted by attackers to the scheduler are not random; there are specific requirements regarding the timing of the submission, the amount of data in the job, and its content type. The attacker's ultimate goal is to collect the scheduler's visible output results for each probe job. Therefore, by collecting and analyzing the scheduler's visible output results, the attacker's probe trajectory on the scheduler can be accurately determined. Specifically, firstly, the quantum operating system scheduler acquires several visible output data corresponding to several input tasks (i.e., several jobs submitted by the user to the scheduler), with each visible output data uniquely corresponding to one input task. The visible output data includes elements such as the waiting time (the time between the input task being submitted to the scheduler and being processed by the scheduler), start time (the time the scheduler begins processing the input task), completion time (the time the scheduler completes processing the input task), and task processing result (e.g., whether the scheduler successfully or unsuccessfully processed the input task). Then, the visible output data is filtered and transformed. All elements within the visible output data undergo text semantic recognition and vectorization transformation to obtain the processing representation vector corresponding to each input task. All processing representation vectors are then sequentially arranged and integrated according to the order in which all input tasks were submitted to the scheduler, resulting in a probe trajectory for the scheduler. This probe trajectory can be, but is not limited to, the probe behavior trajectory corresponding to external probe tasks submitted to the scheduler.

[0064] In another embodiment, determining whether there is a risk of leakage in the detection trajectory based on the time-domain changes of the detection trajectory includes:

[0065] The temporal changes of all processing representation vectors under the detection trajectory are compared to obtain several queues of element changes; among them, the elements include the waiting time, start time, completion time and task processing result of each input task;

[0066] Determine whether each element change queue meets the preset fine-grained conditions; if at least one element change queue meets the preset fine-grained conditions, determine that the detection trajectory has a risk of information leakage; otherwise, determine that the detection trajectory does not have a risk of information leakage.

[0067] The probe trajectory corresponds to the probe behavior trajectory of external forces submitting probe tasks to the scheduler. It includes changes in the probe behavior initiated by the external forces on the scheduler throughout the entire probe process. The temporal changes of all processing representation vectors under the probe trajectory are compared to obtain several element change queues. Each element change queue corresponds to the change of any one of the following over time: waiting time, start time, completion time, or task processing result for each input task. Then, it is determined whether each element change queue meets a preset fine-grained condition. This preset fine-grained condition may include, but is not limited to, whether the fine-grainedness of the element change queue is within a preset fine-grained range. If at least one element change queue meets the preset fine-grained condition, the probe trajectory is determined to have a risk of information leakage; otherwise, it is determined that the probe trajectory does not have a risk of information leakage, thereby accurately identifying the probe behavior initiated by the attacker.

[0068] In another embodiment, when there is a risk of leakage, suspicious individuals related to the scheduler are identified based on the scheduler's job submission history, including:

[0069] If there is a risk of leakage, the job submission log will be obtained based on the job submission interface currently connected to the scheduler.

[0070] The job submission logs are filtered by historical time intervals to obtain historical job submission information. The historical job submission information is then analyzed to identify time-domain changes in job submission status, and users who meet the scheduler's job submission conditions are identified as suspicious users of the scheduler. The job submission status includes the type of job submitted, time distribution, and user identity.

[0071] When there is a risk of information leakage in the probe trajectory, the probe trajectory is initiated by a specific external user. In this case, based on the job submission interface currently connected to the scheduler (i.e., the interface through which all external users submit jobs to the scheduler), the job submission logs generated by the scheduler through the aforementioned job submission interface are obtained. According to a preset historical time interval, the job submission history information corresponding to the preset historical time interval is extracted from the job submission logs, and the job submission status time-domain change identification is performed on the aforementioned job submission history information to identify users who meet the scheduler's job submission conditions. These users are thus identified as suspicious individuals of the scheduler, providing a basis for subsequent attacker identification. The aforementioned job submission conditions may include, but are not limited to, a preset interval regularity in job submission time.

[0072] In another embodiment, the process of tracing the job submissions of suspicious individuals to determine whether they are attackers includes:

[0073] Obtain information on the job submission behavior of suspicious individuals at the job submission interface; the job submission behavior information includes the time information of each job submission by the suspicious individual;

[0074] Based on the job submission behavior information, determine whether the jobs submitted by the suspicious person in the past have caused service delays for the scheduler; if so, determine that the suspicious person is an attacker; otherwise, determine that the suspicious person is not an attacker.

[0075] Understandably, attackers submit jobs to the scheduler not to actually process the jobs, but to obtain a visible representation of the scheduler's processing results. Therefore, attacker-submitted jobs are significantly different from normal job submissions. By obtaining the time information of each job submission by a suspicious individual at the job submission interface, and based on this time information, it can be determined whether the suspicious individual's past job submissions caused service delays to the scheduler. This helps determine if the suspicious individual is an attacker, providing a basis for subsequent defenses to prevent the attacker's probing activities.

[0076] In another embodiment, the attacker's interface interaction logs on the quantum operating system are monitored, and the attacker's job submission characteristics are determined based on the interface interaction logs, including:

[0077] Based on the attacker's identity information, the attacker's interface interaction logs are extracted from the quantum operating system; the interface interaction logs include records of the attacker's interaction behavior with the quantum operating system's submission interface.

[0078] Extract the temporal and content characteristics of the attacker's job submissions to the submission interface from the interface interaction logs, and use these characteristics as job submission features. The temporal and content characteristics include the time point and the amount of job data for each job submission initiated by the attacker.

[0079] In another embodiment, a defense strategy against attackers is determined based on job submission characteristics, including:

[0080] Based on time domain and content characteristics, the occupancy status of the quantum computing side channel of the quantum operating system by the job submitted by the attacker is determined; wherein, the occupancy status includes the time interval of the attacker's occupancy of the channel and the type of resources occupied;

[0081] Based on the occupancy status, determine the channel's defense strategy against attackers; the defense strategy includes strategies to restrict the channel's use by attackers.

[0082] In practice, based on the attacker's identity information, the attacker's interface interaction logs are extracted in the quantum operating system. From these logs, the attacker's job submission characteristics, such as the time of each submission and the amount of job data, are extracted. It can be understood that these job submission characteristics quantitatively reflect the attacker's probing behavior on the scheduler, providing a basis for subsequent execution of defense strategies against the attacker.

[0083] To prevent and restrict attackers' probing behavior on the scheduler and to ensure that other users can use the scheduler normally, based on the aforementioned time domain and content characteristics, the occupation status of the quantum computing side channel of the quantum operating system by the jobs submitted by attackers is determined. The time interval and resource type occupied by the channel when the attacker initiates probing behavior on the scheduler are accurately determined. Then, the channel's defense strategy against attackers is determined, thereby restricting the time interval and resource type occupied by attackers on the channel, and precisely regulating and restricting the behavior of attackers.

[0084] In another embodiment, the job interception operation of the quantum computing side channel is adjusted according to the defense strategy, including:

[0085] Based on the channel restriction policy included in the defense strategy, the interception operation parameters of the quantum computing side channel for the jobs submitted by the attacker are adjusted; wherein, the interception operation parameters include the duration of the interception operation and the type and quantity of resources that are prohibited from being allocated corresponding to the interception operation.

[0086] In another embodiment, the method further includes adjusting the resource configuration of the channel based on the execution status of the job interception operation, which includes:

[0087] Obtain the execution progress of the job interception operation to determine whether the job interception operation has been completed;

[0088] Once the job interception operation is completed, the resource configuration status of the channel for other jobs is adjusted; the resource configuration status includes the type and quantity of resources allocated to other jobs.

[0089] If the job interception operation is not completed, the current interception operation parameters for the channel will remain unchanged.

[0090] In practice, to accurately and promptly execute defense strategies, the duration of the quantum computing side channel's interception operation on attacker-submitted jobs and the types and quantities of resources prohibited from allocation corresponding to the interception operation are adjusted based on the channel restriction policies included in the defense strategy. This prevents the scheduler from being subjected to prolonged probe interference from attackers. Furthermore, the execution progress of the job interception operation is acquired to determine whether the interception operation against the attacker's job has been completed. This allows for adjustments to the channel's resource configuration for other jobs, or maintaining the current interception operation parameters unchanged, precisely defending against attacker snooping, preventing scheduler information leakage, and ensuring scheduler security.

[0091] An embodiment of the quantum computing side-channel defense method based on scheduler output observability control provided by this invention is as follows:

[0092] The "quantum operating system" of this invention may include, but is not limited to, a quantum cloud service control plane, a quantum runtime layer, a quantum job scheduler, a quantum backend resource manager, or a combination thereof. The "scheduler visible output" of this invention refers to information that the job submitting entity can directly or indirectly obtain from the scheduler response, job status query, execution result return, or service interface, including but not limited to job status, timing information, backend selection information, and resource category information. The "output observability control" of this invention refers to controlling the granularity, timing, category, and query frequency of externally visible output without altering the scheduler's internal records.

[0093] Please see Figure 2 The specific embodiments described above include:

[0094] S101, Obtain visible output data from the scheduler. When processing quantum computing jobs, the scheduler returns one or more of the following information to the job submitter: job acceptance status, rejection status, waiting time, start time, completion time, completion delay, retry status, delayed execution status, job processing result, backend identifier, candidate backend set, backend resource type, queue position range, or service level response information. The system treats this externally visible data as the object to be controlled, while simultaneously retaining internal real scheduling logs for auditing and scheduling decisions.

[0095] S102, Constructing the Output Observation Trajectory. For each quantum computing job, the system extracts submission features such as job submission time, line width, line depth, number of gates, number of two-qubit gates, declared runtime, declared resource requirements, isolation requirements, or service level, and associates them with the corresponding visible output data to form an output representation vector. Subsequently, multiple output representation vectors are sorted according to job submission time or scheduler response time to form the output observation trajectory.

[0096] S103, determine the risk of scheduler output observability leakage. The system calculates at least one of the following based on the output observation trajectory: time-series fine-grained metric, backend distinguishability metric, output stability metric, and probe adaptability metric. The time-series fine-grained metric characterizes whether external entities can distinguish hidden scheduling states based on waiting time, start time, or completion delay; the backend distinguishability metric characterizes the degree to which the backend identifier or candidate backend set exposes resource pressure or protected jobs; the probe adaptability metric characterizes whether the job submitting entity changes subsequent submission time, job size, or resource declaration based on historical responses.

[0097] S104, Select the output observability control strategy. When the leakage risk score exceeds the preset risk threshold, or any indicator meets the preset abnormal conditions, the system selects at least one of the following strategies: time-series output coarsening strategy, backend selection masking strategy, and probe response suppression strategy. Different strategies can be used individually or in combination according to tenant-level risk status, service level, and protected operation attributes.

[0098] S105, Execute defense and provide feedback updates. The system adjusts the visible output returned by the scheduler to the job submission entity according to the selected strategy, and updates the admission control parameters, limit control parameters, or output granularity control parameters based on the adjustment results. The system also records the visible output before and after the adjustment, risk score, defense trigger conditions, and the strategy adopted for auditing and subsequent threshold updates.

[0099] The methods for determining the risk of leakage can be as follows:

[0100] In one alternative implementation, the system generates a leakage risk score as follows:

[0101] R = α1*Gt + α2*Gb + α3*Gs + α4*Ga, where R represents the leakage risk score, Gt represents the time-series fine-grained index, Gb represents the back-end distinguishability index, Gs represents the output stability index, Ga represents the detector adaptability index, and α1 to α4 are non-negative weights. The above formula is merely an example; those skilled in the art can also determine leakage risk using rule-based thresholds, classifiers, statistical tests, or information gain methods.

[0102] For example, when the same job submitter submits multiple probe jobs of similar size but different submission times within a short period of time, and the waiting time or completion delay returned by the scheduler shows a stable and distinguishable change, the system can improve Gt and Ga; when the specific backend identifier or candidate backend set is highly correlated with the hidden resource pressure, the system can improve Gb; when the output differences obtained from multiple probes are small and easy to model, the system can improve Gs.

[0103] In another alternative implementation, the system categorizes leakage risks into low, medium, and high risks. For low-risk scenarios, normal output is retained; for medium-risk scenarios, timing output coarsening or backend selection masking is enabled; for high-risk scenarios, probe response suppression, query limit, cooldown time, or increased resource claim verification level are further enabled.

[0104] For timing output coarsening strategies, the following can be used:

[0105] Timing output coarsening is used to reduce fine-grained information in waiting time, start time, completion time, or completion delay. The system can map precise time to preset time buckets, such as mapping second-level or millisecond-level waiting time to intervals such as "0 to 1 minute", "1 to 5 minutes", or "5 to 15 minutes"; it can also add random perturbations to the time returned to the job submission subject, not exceeding the service level's allowable range; and it can merge multiple job responses into a batch release response, making it difficult for external subjects to recover the scheduler's internal state from a single response.

[0106] Without affecting internal scheduling accuracy, the system only changes the external output format. The scheduler still stores actual start time, completion time, and resource allocation records to support billing, auditing, and troubleshooting. For high-service-level jobs, the system can use smaller time buckets; for entities with probing risks, the system can use larger time buckets or delayed return strategies.

[0107] Regarding the backend's choice of masking strategy, the following applies:

[0108] Backend selection masking is used to reduce the exposure of specific quantum backend identifiers, candidate backend sets, or backend rankings to internal resource states. The system can replace specific quantum backend identifiers with backend types, region categories, capability levels, or resource categories. For example, it can replace specific device names with summary information such as "superconducting quantum backend type A" or "backend category that supports operations with more than ten qubits".

[0109] The system can also choose not to return a specific backend identifier before job execution, but only return a status indicating that the job has been received, queued, or is in progress; after the job is completed, it returns a summary of backend information according to the audit strategy. For multiple backends with equivalent execution capabilities, the system can set a unified alias, or perform hidden, aggregated, or randomized sorting processing on the candidate backend set, thereby reducing the ability of attackers to infer resource pressure or the existence of protected jobs by exploiting changes in backend selection.

[0110] For the detection response suppression strategy, it can be as follows:

[0111] Probe response suppression is used to reduce the benefits of repeatedly submitting probe jobs. The system can identify high-frequency repeated submissions, periodic submissions, adaptive submissions, or declarations of abnormal resource changes by the same job submitter within a preset time window, and set a cooldown time, query limit, coarse-grained status code, or delayed response for that subject.

[0112] For example, if an entity repeatedly submits jobs with similar routing structures but adjusted time intervals within a short period and continuously queries job status or backend information, the system can change the entity's subsequent status query return from precise time to status level, or limit the number of backend information queries to a preset limit. For jobs whose declared resource requirements differ significantly from historical actual execution consumption, the system can increase the verification level or reduce the output granularity, rather than directly identifying the entity as an attacker.

[0113] The above method can reduce the efficiency of probe-type interactions in obtaining scheduler side channel information without excessively affecting normal tenants.

[0114] Please see Figure 3 This invention can be deployed in a multi-tenant quantum cloud environment. Specifically, multiple tenants submit jobs to the quantum operating system scheduler through a submission interface. The scheduler performs admission control, resource allocation, backend selection, and execution control. The output observability control module is located on the scheduler's external response path and is responsible for unified control of timing information, backend selection information, and probe responses.

[0115] In a multi-tenant quantum cloud environment, the system can maintain a tenant-level risk status for each tenant. This tenant-level risk status can be determined by historical submission frequency, query count, job size changes, resource declaration deviations, query behavior regarding backend information, and the degree of overlap with the time window of protected jobs. The system can also set different time bucket lengths, backend category granularity, cooldown times, and query count limits for different service levels.

[0116] When the platform has jobs requiring high isolation, priority jobs, or protected tenant jobs, the system can temporarily increase the output coarsening level within the relevant time window, reduce the return of specific backend selection information, and suppress high-frequency probe responses. The above controls do not require disclosing the existence of protected jobs or altering the internal actual scheduling results; rather, they reduce inferability by controlling externally visible output.

[0117] Please see Figure 4 The output observability control strategy of this invention can select control strategies of different intensities according to the leakage risk score, and convert the original visible output into a controlled visible output. This conversion does not change the actual record inside the scheduler, but only changes the form of external feedback to the job submission subject.

[0118] For example, for precise waiting time, the system can map fine-grained timing information such as "12.3s" to time intervals such as "10s-15s"; for specific backend selection, the system can replace the specific backend identifier with available backend groups, resource categories, or capability levels; for continuous probe responses, the system can use delayed returns, rate limiting, or cooling windows to reduce the attacker's ability to obtain stable and differentiated feedback through multiple queries.

[0119] The aforementioned control strategies can be implemented individually or in combination based on tenant-level risk status, protected job time windows, service level constraints, and historical submission behavior. When the risk score decreases or the cooling-off window ends, the system can revert to a lower-intensity output control strategy, thereby maintaining a balance between security and service availability.

[0120] This invention also provides a quantum computing side-channel defense device based on scheduler output observability control, comprising a data acquisition module, a trajectory construction module, a risk determination module, a strategy determination module, and an output control module. The data acquisition module acquires visible output data from the scheduler; the trajectory construction module constructs an output observation trajectory; the risk determination module determines the risk of scheduler output observability leakage; the strategy determination module selects at least one of a timing output coarsening strategy, a back-end selection masking strategy, and a probe response suppression strategy; and the output control module adjusts the scheduler's externally visible output and updates the scheduler's admission control parameters, limit control parameters, or output granularity control parameters.

[0121] The present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable by the processor; when the processor executes the computer program, it implements the method described in any of the above embodiments.

[0122] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the method described in any of the above embodiments.

[0123] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A quantum computing side-channel defense method based on scheduler output observability control, characterized in that, include: Obtain the visible output data of the scheduler of the quantum operating system, and determine the probe trajectory of the scheduler based on the visible output data; Based on the time-domain changes of the detection trajectory, determine whether there is a risk of leakage in the detection trajectory; When there is a risk of leakage, the suspicious party of the scheduler is identified based on the job submission history information of the scheduler; The job submissions of the suspected individuals are traced to determine whether they are attackers. Monitor the attacker's interface interaction logs on the quantum operating system, and determine the attacker's job submission characteristics based on the interface interaction logs; Based on the job submission characteristics, determine the defense strategy against the attacker; Based on the aforementioned defense strategy, adjust the job interception operation of the quantum computing side channel.

2. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 1, characterized in that: Acquire the visible output data of the scheduler of the quantum operating system, and determine the probe trajectory of the scheduler based on the visible output data, including: Obtain several visible output data of the quantum operating system scheduler for several input tasks, each corresponding to a specific input task; wherein, each visible output data includes the scheduler's waiting time, start time, completion time, and task processing result for each input task. The visible output data are transformed to obtain the processing representation vector of each input task; the processing representation vectors of all input tasks are sequentially arranged and integrated to obtain the detection trajectory of the scheduler.

3. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 2, characterized in that: Based on the time-domain changes of the detection trajectory, determine whether the detection trajectory poses a risk of leakage, including: The temporal changes of all processing representation vectors under the detection trajectory are compared to obtain several element change queues; wherein, the elements include the waiting time, start time, completion time, and task processing result of each input task; Determine whether each element change queue meets a preset fine-grained condition; if at least one element change queue meets the preset fine-grained condition, then determine that the detection trajectory has a risk of information leakage; otherwise, determine that the detection trajectory does not have a risk of information leakage.

4. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 1, characterized in that: When there is a risk of leakage, based on the scheduler's job submission history information, suspicious individuals associated with the scheduler are identified, including: If there is a risk of leakage, the job submission log is obtained according to the job submission interface currently connected to the scheduler; The job submission logs are filtered by historical time intervals to obtain job submission history information; the job submission history information is then analyzed to identify time-domain changes in job submission status, and users who meet the job submission conditions of the scheduler are identified as suspicious users of the scheduler; wherein, the job submission status includes the type of submitted job, time distribution, and user identity.

5. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 4, characterized in that: To trace the job submissions of the suspected individuals and determine whether they are attackers, the following steps are performed: Obtain the job submission behavior information of the suspected individual at the job submission interface; wherein, the job submission behavior information includes the time information of each job submission by the suspected individual; Based on the job submission behavior information, determine whether the jobs submitted by the suspicious person in the past have caused service delays for the scheduler; if so, determine that the suspicious person is an attacker; otherwise, determine that the suspicious person is not an attacker.

6. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 1, characterized in that: By monitoring the attacker's interface interaction logs on the quantum operating system, and based on the interface interaction logs, determining the attacker's job submission characteristics, including: Based on the attacker's identity information, the quantum operating system extracts the attacker's interface interaction logs; wherein, the interface interaction logs include records of the attacker's interaction behavior with the quantum operating system's submission interface; Extract the temporal and content characteristics of the attacker's job submission to the submission interface from the interface interaction log, and use these characteristics as job submission features; wherein, the temporal and content characteristics include the time point and the amount of job data of each job submission initiated by the attacker.

7. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 6, characterized in that: Based on the job submission characteristics, a defense strategy against the attacker is determined, including: Based on the time domain and content characteristics, the occupancy status of the quantum computing side channel of the quantum operating system by the job submitted by the attacker is determined; wherein, the occupancy status includes the time interval of the attacker's occupation of the channel and the type of resources occupied; Based on the occupancy status, a defense strategy for the channel against the attacker is determined; wherein the defense strategy includes a strategy for restricting the attacker's use of the channel.

8. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 1, characterized in that: According to the aforementioned defense strategy, the job interception operation of the quantum computing side channel is adjusted, including: Based on the channel restriction policy for the attacker included in the defense strategy, the interception operation parameters of the quantum computing side channel for the jobs submitted by the attacker are adjusted; wherein, the interception operation parameters include the duration of the interception operation and the type and quantity of resources that are prohibited from being allocated corresponding to the interception operation.

9. The quantum computing side-channel defense method based on scheduler output observability control as described in claim 8, characterized in that: It also includes adjusting the resource configuration of the channel based on the execution status of the job interception operation, which includes: The execution progress of the job interception operation is obtained to determine whether the job interception operation has been completed. Once the job interception operation is completed, the resource configuration status of the channel for other jobs is adjusted; wherein, the resource configuration status includes the resource allocation type and quantity for other jobs; If the interception operation is not completed, the current interception operation parameters for the channel remain unchanged.

Images