Low-damage variable-granularity secure erasure method for solid-state storage

Abstract

The invention discloses a low-damage variable-granularity secure erasing method for solid-state storage, and belongs to the technical field of data storage security. In order to solve the problems of fixed erasure granularity, single strategy and large damage to a medium in the existing solid-state storage secure erasure technology, the method comprises the following steps of: responding to a data erasure request of a user, and obtaining multi-dimensional attribute information of to-be-erased data; based on the multi-dimensional attribute information, the optimal erasure granularity is obtained through dynamic calculation of a weight model; generating a corresponding overwriting control plan; selecting a target template from a plurality of predefined disturbance templates, and performing one-time exclusive or operation on the target template and the original to-be-erased data to generate unrecoverable disturbance data; the disturbance data is written into the target storage unit, an ECC decoding verification mechanism is triggered to verify the erasing effect, and if the ECC decoding verification mechanism judges that erasing succeeds, the erasing state is recorded, and metadata is updated; and otherwise, returning to execute S4 until the verification mechanism judges that the erasing is successful or the maximum number of attempts is reached. The method is used for data storage.

Description

TECHNICAL FIELD

[0001] The present application relates to a low-damage variable-granularity secure erasure method and system for solid-state storage, and belongs to the technical field of data storage security. BACKGROUND

[0002] With the rapid development of information technology, solid state drives (SSD) have been widely used in personal computers, enterprise servers, embedded systems and data centers due to their high speed reading and writing, low power consumption and strong shock resistance. Under the background of increasing attention to data security, how to achieve safe and thorough deletion of sensitive data while ensuring device performance and service life has become one of the key challenges in the design of storage systems. The traditional logical deletion method only removes the index information in the file system, and the actual data is still retained in the physical medium, which poses a risk of malicious recovery and theft. Therefore, it is urgent to study efficient and secure erasure technology for solid-state media.

[0003] Currently, the secure erasure methods for SSD mainly include full-disk erasure, block erasure and overwrite erasure. Although full-disk erasure has high security, it is complex and time-consuming to operate, and is difficult to apply to local data deletion scenarios; block erasure relies on the erasure instructions of Flash storage media, but frequent execution will accelerate the physical wear of NAND Flash and shorten the service life of the device; the overwrite erasure method covers the original information by writing random data or specific pattern data to the target storage unit, thereby achieving data irrecoverability. However, the existing overwrite method uses fixed granularity (such as page level or block level) operation, and does not consider multiple factors such as data sensitivity, cell wear, access frequency, etc., which may lead to resource waste or incomplete erasure.

[0004] In addition, due to the write mechanism of NAND Flash, which is "erased first and then written", frequent erasure operations not only affect system performance, but also cause write amplification effect, which aggravates the physical damage of Flash media. Current research attempts to introduce disturbance algorithms, ECC redundancy zone interference and other methods to improve erasure effect, but it is still difficult to achieve effective balance between security and device durability.

[0005] In summary, the existing secure erasure technology for solid-state storage has problems such as fixed erasure granularity, single strategy, and large damage to the medium. Therefore, there is an urgent need for a data erasure method and system that can dynamically adjust the erasure granularity according to data sensitivity, access frequency, physical wear and remaining space ratio, and combine disturbance mechanism and real-time verification to achieve low-damage and high-security data erasure, in order to meet the comprehensive needs of security, efficiency and reliability of data deletion in multiple scenarios. SUMMARY

[0006] The application aims to solve the problems of fixed erasure granularity, single strategy and large damage to the medium in the existing solid-state storage secure erasure technology, and provides a low-damage variable granularity secure erasure method and system for solid-state storage.

[0007] The low-damage variable granularity secure erasure method for solid-state storage comprises the following steps:

[0008] S1, in response to a data erasure request of a user, obtaining multi-dimensional attribute information of data to be erased; the multi-dimensional attribute information comprises data sensitivity, access frequency, physical unit wear degree and storage medium remaining space ratio;

[0009] S2, based on the multi-dimensional attribute information, obtaining an optimal erasure granularity through dynamic calculation by a weight model; the erasure granularity comprises block level, page level, sub-page level and ECC redundancy area level;

[0010] S3, generating a corresponding overwrite control plan according to the optimal erasure granularity obtained in S2;

[0011] S4, based on the overwrite control plan in S3, selecting a target template from a plurality of predefined disturbance templates, and performing one-time XOR operation with original data with erasure to generate un-restorable disturbance data;

[0012] S5, writing the disturbance data in S4 into a target storage unit, triggering an ECC decoding verification mechanism to verify the erasure effect: if the ECC decoding verification mechanism determines that the erasure is successful, recording the erasure state and updating the metadata; otherwise, returning to execute S4 until the verification mechanism determines that the erasure is successful or the maximum number of attempts is reached.

[0013] Preferably, the method for obtaining the data sensitivity in S1 comprises:

[0014] A pre-trained natural language processing model is used to perform semantic analysis on the text data to be erased, and the text data to be erased is converted into a feature vector representing the sensitivity level;

[0015] Specifically, it comprises:

[0016] The target text data is subjected to word segmentation processing to obtain word groups related to the text content, and the obtained word groups are mapped into a high-dimensional vector space to generate corresponding word vectors;

[0017] The word vectors are input into a pre-trained deep bidirectional transformer model for context encoding;

[0018] The encoded word vectors are aggregated to obtain a feature vector corresponding to the data sensitivity.

[0019] Preferably, the specific method for obtaining the optimal erasure granularity through dynamic calculation in S2 comprises:

[0020] Construct the erase granularity selection function:

[0021] ;

[0022] in, Indicates data sensitivity. Indicates access frequency. Indicates the wear degree of the physical unit. Indicates the ratio of remaining space;

[0023] function value Input the preset threshold mapping function and output the corresponding optimal erasure granularity level.

[0024] Preferably, the specific method for selecting the target template from a plurality of predefined perturbation templates as described in S4 includes:

[0025] Calculate the value of the perturbation template selection function based on the page address, timestamp, and system key;

[0026] Based on the value of the selection function, the target template to be used is determined from the address-driven perturbation template, time-driven perturbation template, key-driven pseudo-random perturbation template, and structure perturbation variant template;

[0027] The target template is expanded to a bit length consistent with the target storage unit size and used as a random mask, which is then XORed with the original data.

[0028] Preferably, the perturbation template includes:

[0029] The address-driven perturbation template calculates the hash value and generates perturbation data based on the physical address of the target storage unit.

[0030] Time-driven perturbation template, generating perturbation data based on the current timestamp;

[0031] Key-driven pseudo-random perturbation templates generate perturbation data based on the system key and a random number generator;

[0032] The structural perturbation variant template generates perturbation data by performing bit rotation or bit rearrangement on address-driven perturbation templates, time-driven perturbation templates, or key-driven pseudo-random perturbation templates.

[0033] Preferably, before generating the unrecoverable perturbation data as described in S4, the method further includes:

[0034] Calculate the attention score of the feature vector of the data to be erased;

[0035] Random noise signals are generated using the Laplace algorithm;

[0036] Multiply the intensity value of the generated noise signal with the attention score to obtain noise of each dimension-reduced feature vector;

[0037] Add the noise to the dimension-reduced feature vector to obtain a feature vector containing noise;

[0038] The feature vector containing noise is subjected to one-time XOR operation again.

[0039] Preferably, the ECC decoding verification mechanism of S5 specifically comprises:

[0040] After writing the disturbance data, the data of the target storage unit is read and sent to the ECC decoder to trigger the ECC decoding logic;

[0041] The number of error bits in the decoding process is counted:

[0042] If the number of error bits exceeds the preset correction capability threshold, it is determined that the data is unrecoverable and the erasure is successful;

[0043] If the number of error bits does not exceed the preset correction capability threshold, it is determined that the erasure fails.

[0044] Preferably, the specific method for updating the metadata of S5 comprises:

[0045] The wear level count of the physical unit corresponding to the erased data is reduced;

[0046] The storage space of the physical unit is released, and the remaining space ratio is updated;

[0047] The state of the physical unit is marked as erased available in the metadata table.

[0048] The low-damage variable-granularity safe erasure system for solid-state storage comprises:

[0049] An attribute analysis module is configured to acquire and analyze multi-dimensional attribute information of data to be erased;

[0050] A granularity decision module is configured to dynamically calculate and select an optimal erasure granularity based on the multi-dimensional attribute information through a weight model;

[0051] A plan generation module is configured to generate an overwrite control plan according to the selected erasure granularity;

[0052] A disturbance overwrite module is configured to select a disturbance template according to the overwrite control plan, generate disturbance data, and perform a write operation;

[0053] A verification feedback module is configured to trigger an ECC decoding verification mechanism and update an erasure state according to a verification result.

[0054] Preferably, the erasing system supports multi-level erasing strategy switching function, and can dynamically adjust the priority of the erasing strategy according to the real-time running state of the system.

[0055] The low-damage variable-granularity secure erasing method and the erasing system for solid-state storage effectively avoid the time-consuming problem of traditional full-disk erasing and the life consumption problem of frequent block erasing. Through fine-grained control and one-time disturbance overwrite operation, the system response speed and processing efficiency are improved while the security is ensured. The method has the following advantages:

[0056] 1. Dynamic adjustment of secure erasing granularity, achieving a balance between precise security and low damage: The present application breaks through the limitation of fixed erasing granularity in the prior art, dynamically selects the optimal erasing granularity of block level, page level, sub-page level or ECC area by constructing a weight model based on multi-dimensional factors such as data sensitivity, access frequency, physical wear degree and residual space ratio. This ensures that the "minimum necessary writing" is realized under the premise of meeting the data unrecoverable security requirements, effectively reduces the write amplification coefficient, significantly reduces the physical damage to the NAND Flash medium caused by erasing operation, and prolongs the service life of the solid-state storage device.

[0057] 2. Introducing multiple disturbance mechanisms greatly improves data unrecoverability and anti-analysis ability: By integrating address-driven, time-driven, key-driven pseudo-random and structure variant disturbance four templates, the data is subjected to one-time XOR disturbance overwrite. This multi-strategy fusion disturbance method not only ensures that the original data is completely destroyed and cannot be recovered through hardware means, but also effectively resists differential attacks based on storage mode analysis, greatly enhancing the physical layer data security.

[0058] 3. Real-time, reliable and unforgeable erasing verification mechanism: The present application innovatively uses the inherent ECC decoding logic of the storage device as a verification means. By actively triggering ECC decoding and checking whether it fails, it instantly confirms whether the data is unrecoverable. This mechanism is endogenous to the storage hardware, and the verification result is real and reliable and difficult to forge, providing users with intuitive and efficient secure erasing feedback and meeting high-standard security audit requirements. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 is the principle diagram of the low-damage variable-granularity secure erasing method for solid-state storage. DETAILED DESCRIPTION

[0060] With reference to the drawings and specific embodiments, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of the present application.

[0061] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0062] The present application will be further described below in combination with the drawings and specific embodiments, but is not limited to the embodiments.

[0063] Embodiment 1:

[0064] The present application will be further described below in combination with the drawings and specific embodiments, but is not limited to the embodiments. Figure 1 The present application will be further described below in combination with the drawings and specific embodiments, but is not limited to the embodiments.

[0065] S1, in response to a data erasure request of a user, obtaining multi-dimensional attribute information of data to be erased; the multi-dimensional attribute information includes data sensitivity, access frequency, physical unit wear degree and storage medium remaining space ratio;

[0066] S2, based on the multi-dimensional attribute information, obtaining an optimal erasure granularity by dynamic calculation through a weight model; the erasure granularity includes block level, page level, sub-page level and ECC redundancy area level;

[0067] S3, generating a corresponding overwrite control plan according to the optimal erasure granularity obtained in S2;

[0068] S4, based on the overwrite control plan in S3, selecting a target template from a plurality of predefined disturbance templates, and performing a one-time XOR operation with original data with erasure to generate unresumable disturbance data;

[0069] S5, writing the disturbance data in S4 into a target storage unit, triggering an ECC decoding verification mechanism to verify the erasure effect: if the ECC decoding verification mechanism determines that the erasure is successful, recording the erasure state and updating the metadata; otherwise, returning to execute S4 until the verification mechanism determines that the erasure is successful or the maximum number of attempts is reached.

[0070] Further, the method for obtaining the data sensitivity in S1 comprises:

[0071] adopting a pre-trained natural language processing model to perform semantic analysis on the text data to be erased, and converting the text data to be erased into a feature vector representing a sensitive level;

[0072] Specifically, it comprises:

[0073] The target text data is segmented to obtain word groups related to the text content, and the obtained word groups are mapped to a high-dimensional vector space to generate corresponding word vectors;

[0074] The word vectors are input into a pre-trained deep bidirectional transformer model for context encoding;

[0075] The encoded word vectors are aggregated to obtain a feature vector corresponding to the data sensitivity.

[0076] Further, the specific method of dynamically calculating the optimal erasure granularity in S2 includes:

[0077] An erasure granularity selection function is constructed:

[0078] ;

[0079] wherein, represents the data sensitivity, represents the access frequency, represents the physical unit wear degree, represents the remaining space ratio;

[0080] The function value is input into a preset threshold mapping function, and the corresponding optimal erasure granularity level is output.

[0081] Further, the specific method of selecting a target template from a plurality of predefined disturbance templates in S4 includes:

[0082] According to the page address, the timestamp and the system key, the value of the disturbance template selection function is calculated;

[0083] According to the value of the selection function, the target template to be used is determined from the address-driven disturbance template, the time-driven disturbance template, the key-driven pseudo-random disturbance template and the structure disturbance variant template;

[0084] The target template is expanded to a bit length consistent with the size of the target storage unit, serving as a random mask, and is XORed with the original data.

[0085] Further, the disturbance template includes:

[0086] The address-driven disturbance template calculates a hash value based on the physical address of the target storage unit to generate disturbance data;

[0087] The time-driven disturbance template generates disturbance data based on the current timestamp;

[0088] The key-driven pseudo-random disturbance template generates disturbance data based on the system key and a random number generator;

[0089] The structure disturbance variant template generates the disturbed data by performing bit rotation or bit rearrangement on the address-driven disturbance template, the time-driven disturbance template, or the key-driven pseudo-random disturbance template.

[0090] Further, before the step S4 of generating the un-recoverable disturbed data, the method further comprises:

[0091] calculating an attention score of the feature vector of the data to be erased;

[0092] generating a random noise signal by using a Laplace algorithm;

[0093] multiplying the intensity value of the generated noise signal with the attention score to obtain noise of each reduced-dimension feature vector;

[0094] adding the noise to the reduced-dimension feature vector to obtain a feature vector containing noise;

[0095] performing one-time XOR operation on the feature vector containing noise.

[0096] Further, the ECC decoding verification mechanism of the step S5 specifically comprises:

[0097] after the disturbed data is written, reading data of the target storage unit and sending the data into an ECC decoder to trigger ECC decoding logic;

[0098] counting the number of error bits in the decoding process;

[0099] if the number of error bits exceeds a preset correction capability threshold, determining that the data is un-recoverable and the erasing is successful;

[0100] if the number of error bits does not exceed the preset correction capability threshold, determining that the erasing fails.

[0101] Further, the specific method of updating the metadata of the step S5 comprises:

[0102] reducing a wear level count of a physical unit corresponding to the erased data;

[0103] releasing a storage space of the physical unit and updating a remaining space ratio;

[0104] marking a state of the physical unit as erased available in a metadata table.

[0105] Embodiment 2

[0106] The following will be described in combination with Figure 1 the present embodiment. The solid-state storage-oriented low-damage variable-granularity secure erasing system of the present embodiment comprises:

[0107] an attribute analysis module configured to acquire and analyze multi-dimensional attribute information of data to be erased;

[0108] a granularity decision module, configured to dynamically calculate and select an optimal erasing granularity based on the multi-dimensional attribute information through a weight model;

[0109] a plan generation module, configured to generate an overwrite control plan according to the selected erasing granularity;

[0110] a disturbance overwrite module, configured to select a disturbance template according to the overwrite control plan, generate disturbance data and perform a write operation;

[0111] a verification feedback module, configured to trigger an ECC decoding verification mechanism and update an erasing state according to a verification result.

[0112] Further, the erasing system supports a multi-level erasing strategy switching function and can dynamically adjust the priority of the erasing strategy according to a real-time running state of the system.

[0113] In the present application, in order to solve the problems of high write amplification, high energy consumption, fast life consumption and fixed erasing granularity of the existing safe erasing means in solid-state storage devices, a data un-restorable erasing method with high efficiency and low damage is proposed, which can dynamically adjust the erasing granularity according to the data sensitivity and the system state and combine the structure disturbance and the ECC verification mechanism.

[0114] Firstly, the system extracts four key indicators related to the target data after receiving the erasing request: data sensitivity , access frequency , physical unit wear degree and remaining space ratio . The above indicators are used as inputs to construct an erasing granularity selection function , and the mathematical expression is as follows:

[0115] ;

[0116] wherein, , , and are weight coefficients set empirically, and satisfy: .

[0117] According to the interval where the function value is located, the system automatically maps and selects the corresponding erasing granularity level: if , block-level erasing is selected; if , page-level erasing is selected; if , sub-page-level erasing is selected; and if , ECC redundancy area erasing is selected.

[0118] Subsequently, the system generates an erase plan according to the selected granularity, and calls a disturbance template generation module. The selection of the disturbance template is based on the combination of the page address, the timestamp and the system key, and the use of which type of template is determined by a disturbance selection function: including an address-driven template, a time-driven template, a key-driven pseudo-random template and a structure disturbance variant template. The address-driven disturbance template is generated by hashing the page address (PageID), ensuring that each page corresponds to a unique disturbance pattern:

[0119]

[0120] wherein is a lightweight hash function (such as SipHash), and the output is expanded to the page size (such as 4KB) as the disturbance template. Since the page address is unique in the system, it has good non-repeatability and spatial distribution.

[0121] The time-driven disturbance template uses the timestamp of the write time as the disturbance factor to generate a time-related disturbance:

[0122]

[0123] This template can prevent attackers from inferring the disturbance rule through static analysis, and enhance the time sequence unpredictability of the disturbance.

[0124] The key-driven pseudo-random disturbance template uses a pseudo-random number generator (such as AES-CTR, ChaCha20) to generate a disturbance template with the system key as the seed:

[0125]

[0126] wherein Key is the system set key, and Nonce can be generated by combining the page address and the timestamp to ensure that each disturbance is unique and unpredictable.

[0127] The structure disturbance variant template performs bit rotation or rearrangement operations based on existing templates (such as T1 or T2):

[0128]

[0129] The generated disturbance template will be expanded to the same bit length as the original data and perform a one-time XOR operation with the original data to generate a disturbed data page:

[0130]

[0131] The resulting disturbed data is written to the target storage unit, and this process does not require multiple rounds of writing, avoiding the write amplification and media wear problems caused by traditional multiple erasures.

[0132] After the data is overwritten and written, the system immediately triggers the ECC decoding module to decode the page data. If the ECC decoding fails, it indicates that the original data has been completely destroyed and cannot be recovered, and the system will record the successful erasing state and update the logical mapping table and the metadata structure. If the decoding is successful, the system automatically reselects the disturbance template and performs the next round of erasing action until the preset maximum number of rounds is reached or the erasing is successful. This mechanism uses the existing ECC hardware logic in the NAND Flash controller, without additional hardware overhead, and has the advantages of strong real-time performance, unforgeability and high reliability.

[0133] To further improve the security of data erasing and the unpredictability of disturbance, the application introduces a dynamic evolution mechanism of the disturbance template. In each erasing process, the generation of the disturbance template depends on the unique identifier (such as PageID) of the current page, the system timestamp and the dynamic key seed, ensuring that even if the attacker obtains part of the disturbance data, the original data cannot be recovered through differential analysis or template backtracking means. In addition, the structure disturbance template significantly improves the complexity of the disturbance space through lightweight operations such as bit rotation and byte rearrangement, while maintaining low computational resource consumption, suitable for embedded SSD controller environment.

[0134] Finally, the application also supports a multi-level policy switching mechanism. In the idle state of the device, a fine-grained erasing strategy can be used to reduce write amplification; in high-load or high-frequency erasing scenarios, the system can automatically switch to a page-level or block-level strategy to improve throughput efficiency. This mechanism dynamically adjusts the priority of the erasing strategy by monitoring the system load and storage resource state in real time, ensuring an optimal balance between security, performance and lifespan.

[0135] In the present application, a data security erasing method and system for solid state storage devices are disclosed, especially suitable for solid state drives (SSD) and other storage media based on NAND Flash. The method constructs a variable-granularity overwrite model, dynamically selects erasing granularity (including block-level, page-level, sub-page-level and ECC area) according to factors such as data sensitivity, access frequency, physical wear degree and remaining space ratio, to achieve data irrecoverability with the minimum necessary write amount. Further, combined with four types of structure disturbance templates (address-driven, time-driven, key-driven pseudo-random, and structure variant disturbance), the data is subjected to one-time XOR disturbance overwrite, improving the physical layer security and resisting differential analysis attacks. The method also introduces a real-time verification mechanism based on ECC decoding failure, realizing immediate feedback and unforgeable verification of erasing effect. The application meets the security requirements of NISTSP 800-88 Rev.1, effectively reduces the write amplification coefficient, prolongs the device lifespan, and is suitable for solid state storage systems with high requirements for data security and device reliability.

[0136] While the application has been described with reference to particular embodiments, it is to be understood that the application is not limited to the particulars disclosed. Rather, it is a continuation of the principles and applications of the present application. It is therefore to be understood that numerous modifications, both as to the details and embodiments illustrated and the application defined by the appended claims, can be made as forms of the application. It is also to be understood that the terminology used herein is for the purpose of describing the particular embodiments only and is not intended to limit the scope of the present application which will be limited only by the appended claims. It is further understood that the features described in connection with one embodiment can be used in conjunction with another embodiment.

Claims

1. A low-damage, variable-granularity secure erasure method for solid-state storage, characterized in that, It includes: S1. Respond to the user's data erasure request and obtain the multi-dimensional attribute information of the data to be erased; The multidimensional attribute information includes data sensitivity, access frequency, physical unit wear and tear, and the remaining space ratio of the storage medium; S2. Based on multi-dimensional attribute information, the optimal erasure granularity is dynamically calculated through a weight model; the erasure granularity includes block level, page level, subpage level, and ECC redundancy area level. S3. Generate the corresponding overwrite control plan based on the optimal erase granularity obtained in S2; S4. Based on the overwrite control plan described in S3, select the target template from a number of predefined disturbance templates, perform a one-time XOR operation with the original erased data, and generate unrecoverable disturbance data. S5. Write the disturbance data described in S4 into the target storage unit and trigger the ECC decoding verification mechanism to verify the erasure effect: If the ECC decoding verification mechanism determines that the erasure is successful, record the erasure status and update the metadata; otherwise, return to execute S4 until the verification mechanism determines that the erasure is successful or the maximum number of attempts is reached.

2. The low-damage, variable-granularity secure erasure method for solid-state storage according to claim 1, characterized in that, The methods for obtaining the data sensitivity described in S1 include: A pre-trained natural language processing model is used to perform semantic analysis on the text data to be erased, and the text data to be erased is converted into feature vectors representing the sensitivity level. Specifically, it includes: The target text data is segmented to obtain word groups related to the text content. The obtained word groups are then mapped to a high-dimensional vector space to generate corresponding word vectors. The word vectors are input into a pre-trained deep bidirectional transformer model for context encoding; The encoded word vectors are aggregated to obtain the feature vector corresponding to the data sensitivity.

3. The low-damage, variable-granularity secure erasure method for solid-state storage according to claim 1, characterized in that, The specific method for obtaining the optimal erasure granularity through dynamic calculation as described in S2 includes: Construct the erase granularity selection function: ； in, Indicates data sensitivity. Indicates access frequency. Indicates the wear degree of the physical unit. Indicates the ratio of remaining space; function value Input the preset threshold mapping function and output the corresponding optimal erasure granularity level.

4. The low-damage, variable-granularity secure erasure method for solid-state storage according to claim 1, characterized in that, The specific method for selecting a target template from a plurality of predefined perturbation templates, as described in S4, includes: Calculate the value of the perturbation template selection function based on the page address, timestamp, and system key; Based on the value of the selection function, the target template to be used is determined from the address-driven perturbation template, time-driven perturbation template, key-driven pseudo-random perturbation template, and structure perturbation variant template; The target template is expanded to a bit length consistent with the target storage unit size and used as a random mask, which is then XORed with the original data.

5. The low-damage, variable-granularity secure erasure method for solid-state storage according to claim 4, characterized in that, The disturbance template includes: The address-driven perturbation template calculates the hash value and generates perturbation data based on the physical address of the target storage unit. Time-driven perturbation template, generating perturbation data based on the current timestamp; Key-driven pseudo-random perturbation templates generate perturbation data based on the system key and a random number generator; The structural perturbation variant template generates perturbation data by performing bit rotation or bit rearrangement on address-driven perturbation templates, time-driven perturbation templates, or key-driven pseudo-random perturbation templates.

6. The low-damage, variable-granularity secure erasure method for solid-state storage according to claim 1, characterized in that, Before generating the unrecoverable disturbance data as described in S4, the following is also included: Calculate the attention score of the feature vector of the data to be erased; Random noise signals are generated using the Laplace algorithm; The intensity value of the generated noise signal is multiplied by the attention score to obtain the noise of each dimension-reduced feature vector; The noise is added to the dimensionality-reduced feature vector to obtain a feature vector containing the noise. Then perform a one-time XOR operation on the feature vector containing noise.

7. The low-damage, variable-granularity secure erasure method for solid-state storage according to claim 1, characterized in that, The ECC decoding verification mechanism described in S5 specifically includes: After writing the perturbation data, the data in the target storage unit is read and sent to the ECC decoder, triggering the ECC decoding logic; Count the number of erroneous bits during the decoding process: If the number of error bits exceeds the preset correction capability threshold, the data is determined to be unrecoverable and the erasure is successful. If the number of error bits does not exceed the preset correction capability threshold, the erasure is deemed to have failed.

8. The low-damage, variable-granularity secure erasure system for solid-state storage according to claim 1, characterized in that, The specific methods for updating metadata as described in S5 include: Reduce the wear count of the physical cells corresponding to the erased data; Release the storage space of the physical unit and update the remaining space ratio; The physical unit's status is marked as erased and available in the metadata table.

9. An erasure system implementing the low-damage variable-granularity secure erasure method for solid-state storage as described in any one of claims 1-8, characterized in that, It includes: The attribute analysis module is used to acquire and analyze the multidimensional attribute information of the data to be erased; The granularity decision module is used to dynamically calculate and select the optimal erasure granularity based on the multidimensional attribute information and through a weight model. The plan generation module is used to generate an overwrite control plan based on the selected erase granularity; The disturbance overwrite module is used to select a disturbance template according to the overwrite control plan, generate disturbance data and perform a write operation; The verification feedback module is used to trigger the ECC decoding verification mechanism and update the erasure status based on the verification result.

10. The low-damage, variable-granularity secure erasure system for solid-state storage according to claim 9, characterized in that, The erasure system supports multi-level erasure strategy switching, and can dynamically adjust the priority of the erasure strategy according to the real-time operating status of the system.

Images