An adaptive encryption method and system based on semantic features of mail content

Abstract

The application discloses a kind of based on mail content semantic feature self-adapting encryption method and system, it is related to natural language processing technical field, the method includes: receiving to be encrypted mail, the semantic role labeling of mail content is extracted key entity and its semantic category;Obtain the sentiment intensity coefficient of key entity, and judge the time sensitivity of each key entity, generate time urgency coefficient;Based on the semantic category of key entity, sentiment intensity coefficient and time urgency coefficient, construct three-dimensional evaluation matrix, calculate comprehensive encryption sensitivity;Based on the combination logic of entity category, sentiment coefficient and time coefficient, preset encryption rule is matched, and adaptive encryption strategy containing encryption intensity and encryption effective period is dynamically generated;Based on adaptive encryption strategy, the semantic paragraph division of mail content is carried out, and differential encryption processing is executed to the semantic paragraph containing key entity, and bind time-aware access control.

Description

Technical Field

[0001] This invention relates to the field of natural language processing technology, specifically an adaptive encryption method and system based on the semantic features of email content. Background Technology

[0002] In email communication, encrypting sensitive information is a fundamental requirement for ensuring data security. Traditional email encryption methods mainly fall into two categories: rule-based static encryption, which applies a fixed strength of encryption to all emails or emails matching simple keyword rules (this approach lacks flexibility and can easily lead to insufficient security or wasted performance); and machine learning-based classification encryption, which uses trained models to determine whether an email is sensitive as a whole and thus decides whether to encrypt it. However, in existing technologies, static rule-based methods cannot understand the deep semantics and context of email content, making it difficult to accurately identify the information units that truly need protection; while machine learning methods typically treat emails as a whole, failing to achieve paragraph-level differentiated encryption, and the models rely on large amounts of labeled data, resulting in poor interpretability and difficulty adapting to the complex and ever-changing semantic expressions, sentiments, and time sensitivity in emails. Furthermore, existing solutions generally lack intelligent management of the encryption lifecycle and cannot dynamically adjust encryption strategies based on the time value of the information. Summary of the Invention

[0003] The purpose of this invention is to provide an adaptive encryption method and system based on the semantic features of email content, so as to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an adaptive encryption method based on the semantic features of email content, the method comprising: S100: Receive the email to be encrypted, perform semantic role labeling on the email content, and extract at least one key entity and its semantic category. S200. For each key entity, analyze the sentiment tendency of its textual context and obtain the sentiment intensity coefficient of the key entity. S300: Identify the time semantic information in the email content, determine the time sensitivity of each key entity, and generate a time urgency coefficient; S400: Based on the semantic category, sentiment intensity coefficient, and time urgency coefficient of key entities, a three-dimensional evaluation matrix is ​​constructed to calculate the comprehensive encryption sensitivity. S500: Based on the comprehensive encryption sensitivity, and using the combination logic of entity category, sentiment coefficient and time coefficient, it matches the preset encryption rules and dynamically generates an adaptive encryption strategy that includes encryption strength and encryption validity period. S600, based on an adaptive encryption strategy, divides email content into semantic segments, performs differentiated encryption processing on semantic segments containing key entities, and binds time-aware access control.

[0005] According to the above scheme, step S100 includes: S110. Perform syntactic analysis on the sentences in the email content and identify the predicate verbs in the sentences; S120. Focusing on the predicate verb, analyze and label the semantic roles of other components in the sentence. Semantic roles include agent role, patient role, time, place, and manner. The analysis process categorizes roles based on the dependency relationship between syntactic components and the predicate verb to ensure logical consistency in labeling. S130. Based on the labeled semantic roles, extract nouns or noun phrases as core entities from the text components corresponding to the agent and patient roles, and use them as candidate key entities. The text components include nouns, pronouns, or noun phrases, and their syntactic functions and semantic roles are used to determine whether they are entities that carry core information. S140. Based on a pre-defined entity category thesaurus, perform semantic classification on candidate key entities to determine at least one key entity and its corresponding semantic category. Semantic categories include people, organizations, project names, and proprietary data. The entity category thesaurus contains a multi-level semantic tag system. The classification process is completed by string matching of entity names or semantic similarity calculation to ensure the accuracy of category classification.

[0006] According to the above scheme, step S200 includes: S210. Delineate a context window containing the key entity, centered on the location of the key entity in the email text; the size of the context window is adaptively adjusted according to the semantic density of the text and the entity type to capture the most relevant contextual information. S220. Match sentiment words and sentiment modifiers in the text within the context window to identify sentiment polarity words and sentiment intensity modifiers. The matching process is based on a preset sentiment dictionary and modifier dictionary, while also considering the influence of word negation modification and transition logic on the final sentiment tendency. S230. Based on the combination relationship between the identified sentiment polarity words and sentiment intensity modifier words, calculate the basic sentiment intensity value of the context window; the calculation process is based on the baseline score of the polarity words and the intensity coefficient of the modifier words to obtain a standardized sentiment intensity value. S240. Based on the semantic category of the key entity, assign an entity category weight factor to the basic sentiment intensity value to obtain the sentiment intensity coefficient of the key entity; the entity category weight factor is pre-set according to the sensitivity of the entity to sentiment information, reflecting the risk difference of different categories of entities; According to the above scheme, step S300 includes: S310. Perform time semantic analysis on the email content to identify and extract explicit time expressions and implicit time references; explicit time expressions include specific dates, times or period descriptions, and implicit time references include time information inferred through events, states or logical associations; S320. Establish the association mapping between temporal semantic information and each key entity. By analyzing the semantic connection relationship between time statements and key entities, determine the target time type associated with each key entity. The target time type includes past time, current time and future time. The semantic connection relationship is determined by analyzing the dependency distance, logical modification relationship and co-occurrence relationship between entities and time statements in the syntactic structure. S330. Based on the target time type and combined with the degree of association between time semantic information and key entities, determine the time sensitivity level of each key entity; the degree of association is determined by comprehensively quantifying the syntactic distance, logical dependency strength and semantic coherence between the entity and the time representation. S340. Map the time sensitivity level to a quantitative parameter, and perform weighted adjustment based on the time distance attribute of the target time type associated with the key entity to generate the time urgency coefficient of each key entity; the time distance attribute is calculated based on the difference between the target time type and the current system time, with future time having positive urgency and past time having negative or zero urgency.

[0007] According to the above scheme, step S400 includes: S410. Based on the semantic categories of key entities, map them to the corresponding semantic quantization values; S420. Based on the overall semantic context of the current email, dynamically allocate and adjust the dimensional weight factors of semantic quantification value, emotional intensity coefficient and time urgency coefficient; the overall semantic context is obtained through comprehensive analysis of the email's subject word distribution, stylistic features and writing intention, and the adjustment of the weight factors aims to reflect the differences in the importance of each dimension to the encryption decision under different contexts. S430. Using semantic quantification value, emotional intensity coefficient and time urgency coefficient as coordinates of the three-dimensional evaluation space, and using their respective corresponding dimension weight factors as coefficients, a three-dimensional evaluation matrix is ​​constructed. S440. Perform weighted fusion calculation on the parameters of each dimension in the three-dimensional evaluation matrix to obtain the comprehensive encryption sensitivity corresponding to each key entity; the weighted fusion calculation adopts linear weighted summation or nonlinear function mapping to ensure the reasonable integration of the contribution of each dimension; S450. If an email contains multiple key entities, the maximum value among all comprehensive encryption sensitivity scores shall be selected as the comprehensive encryption sensitivity score of the email.

[0008] According to the above scheme, step S500 includes: S510. Based on the combination relationship of semantic category, sentiment intensity coefficient and time urgency coefficient of key entities, a preset encryption rule base is constructed. The preset encryption rule base defines the encryption strength level and encryption validity period range corresponding to different combination relationships. S520. Based on the overall encryption sensitivity of the current email, and combined with the semantic category, sentiment intensity coefficient and time urgency coefficient of the key entities, match in the preset encryption rule base to locate the target rule entry. S530. Based on the target rule entries and with reference to the quantitative value of the sentiment intensity coefficient, determine the encryption strength level of the current email. The encryption strength level is positively correlated with the information security attribute of the semantic category and the quantitative value of the sentiment intensity coefficient. That is, the higher the inherent risk of the entity and the stronger the sentiment, the stronger the encryption algorithm or key length should be increased accordingly. S540. Based on the target rule entries and referring to the quantification results of the time urgency coefficient, determine the encryption validity period in the adaptive encryption strategy. The encryption validity period is negatively correlated with the quantification value of the time urgency coefficient. That is, the more urgent the time and the stronger the timeliness of the information, the shorter the encryption validity period should be set to achieve automatic expiration protection of information. S550 integrates encryption strength level and encryption validity period to dynamically generate an adaptive encryption strategy for the current email; the generated adaptive encryption strategy is a structured instruction set that specifies the encryption algorithm, key management method and access time window to be used for different paragraphs.

[0009] According to the above scheme, step S600 includes: S610. Based on the semantic relevance of the email content and the distribution of key entities, the email text is divided into semantic paragraphs to ensure that key entities and their closely related semantic content are located in the same semantic paragraph. The division process is based on punctuation marks, paragraph marks and semantic transitions as boundaries, and is adjusted according to the referential relationship between entities and the coherence of the topic. S620. For each semantic segment after division, determine the corresponding segment encryption level according to the semantic category of the key entities contained therein and the encryption strength level determined in the adaptive encryption strategy. The segment encryption level is a refinement and adaptation of the overall encryption strength level of the email to the specific segment, and varies depending on the different entities within the segment. S630. Based on the paragraph encryption level, determine and execute the corresponding paragraph encryption processing method to achieve differentiated encryption for different semantic paragraphs; the encryption processing method includes selecting symmetric encryption algorithms, asymmetric encryption algorithms or hybrid encryption schemes of different strengths, and independently encapsulating the encrypted paragraphs; S640. Based on the encryption validity period determined in the adaptive encryption strategy, establish time-aware access permission rules; the access permission rules are implemented by binding the validity period information with the encryption key or access token through digital signature to ensure the immutability of time constraints. S650. Based on the access permission rules, access control is bound to the encrypted email content, so that the access permission is effective within the encryption validity period and automatically terminates after expiration, thus completing the binding of encryption and access control. After binding, when an access request is made, the current time will be verified to see if it is within the validity period, and the decryption operation will be allowed or denied accordingly, thus realizing fully automatic lifecycle management.

[0010] An adaptive encryption system based on the semantic features of email content, comprising a feature extraction module, a semantic evaluation module, a policy generation module, and a segmented encryption module; The feature extraction module is used to receive emails to be encrypted, perform semantic role labeling on the email content, and extract at least one key entity and its semantic category. The semantic evaluation module is used to analyze the sentiment tendency of the text context in which each key entity is located to obtain the sentiment intensity coefficient, identify the time semantic information in the email content to determine the time sensitivity and generate the time urgency coefficient, and construct a three-dimensional evaluation matrix based on the semantic category, sentiment intensity coefficient and time urgency coefficient of the key entity to calculate the comprehensive encryption sensitivity. The strategy generation module is used to dynamically generate an adaptive encryption strategy that includes encryption strength and encryption validity period by matching preset encryption rules based on the comprehensive encryption sensitivity and the combination logic of entity category, sentiment coefficient and time coefficient. The segmented encryption module is used to divide email content into semantic segments based on an adaptive encryption strategy, perform differentiated encryption processing on semantic segments containing key entities, and bind time-aware access control.

[0011] According to the above scheme, the semantic evaluation module includes a sentiment analysis unit, a time analysis unit, and a fusion evaluation unit; The sentiment analysis unit is used to define a context window centered on key entities, identify sentiment words within the window, and calculate the sentiment intensity coefficient. The time analysis unit is used to identify time semantic information in emails, establish semantic associations between them and key entities, and generate a time urgency coefficient based on time type and the degree of association. The fusion evaluation unit is used to quantify semantic categories into semantic quantification values, dynamically allocate the weights of each dimension according to the overall semantic context of the email, construct a three-dimensional evaluation matrix, perform weighted fusion calculations, and output the comprehensive encryption sensitivity.

[0012] According to the above scheme, the segmented encryption module includes a segment analysis unit and a control binding unit; The paragraph analysis unit is used to divide semantic paragraphs based on the semantic relevance and distribution of key entities in the email, and to determine the paragraph encryption level based on the semantic category of key entities within the paragraph and the encryption strength level in the adaptive encryption strategy. The control binding unit is used to perform differentiated encryption based on the paragraph encryption level, and to establish and bind time-aware access control rules according to the encryption validity period in the adaptive encryption strategy.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention analyzes email content by integrating three-dimensional features of semantic roles, sentiment tendencies, and time sensitivity, enabling a deep understanding of the context and intent of the email content, dynamically assessing the encryption requirements of different content segments, and assigning appropriate encryption strength and validity period to different semantic segments. 2. This invention generates encryption strategies through interpretable semantic analysis logic and explicit combination rules; this makes it easier for administrators to understand and audit the basis for encryption decisions, thereby enhancing credibility and maintainability; 3. This invention quantifies the temporal semantic information in email content and dynamically associates the encryption validity period with the actual time sensitivity of the content, thereby realizing intelligent access control that automatically expires; and achieving automated and precise management of the information lifecycle. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the steps of an adaptive encryption method based on semantic features of email content according to the present invention. Figure 2 This is a schematic diagram of the structure of an adaptive encryption system based on the semantic features of email content according to the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Example: Figures 1-2 As shown, this invention provides a technical solution: an adaptive encryption method based on the semantic features of email content, which includes the following steps: S100: Receive the email to be encrypted, perform semantic role labeling on the email content, and extract at least one key entity and its semantic category. S200. For each key entity, analyze the sentiment tendency of its textual context and obtain the sentiment intensity coefficient of the key entity. S300: Identify the time semantic information in the email content, determine the time sensitivity of each key entity, and generate a time urgency coefficient; S400: Based on the semantic category, sentiment intensity coefficient, and time urgency coefficient of key entities, a three-dimensional evaluation matrix is ​​constructed to calculate the comprehensive encryption sensitivity. S500: Based on the comprehensive encryption sensitivity, and using the combination logic of entity category, sentiment coefficient and time coefficient, it matches the preset encryption rules and dynamically generates an adaptive encryption strategy that includes encryption strength and encryption validity period. S600, based on an adaptive encryption strategy, divides email content into semantic segments, performs differentiated encryption processing on semantic segments containing key entities, and binds time-aware access control.

[0017] In the first embodiment, the content of the email to be encrypted is as follows: "According to the meeting decision, the core architecture design document for Project A must be finalized and submitted for archiving by 18:00 on January 1, 2026. This document involves the company's key future technology roadmap, and any leakage may cause incalculable strategic losses. The marketing department reports that competitor B has been making frequent moves recently, and we need to be vigilant. Project leader C emphasized at last week's meeting that this is a top priority matter." Specifically, step S100 includes: S110. Perform syntactic analysis on the sentences in the email content and identify the predicate verbs in the sentences; S120. Focusing on the predicate verb, analyze and label the semantic roles of other components in the sentence. Semantic roles include agent role, patient role, time, place, and manner. The analysis process categorizes roles based on the dependency relationship between syntactic components and the predicate verb to ensure logical consistency in labeling. S130. Based on the labeled semantic roles, extract nouns or noun phrases as core entities from the text components corresponding to the agent and patient roles, and use them as candidate key entities. The text components include nouns, pronouns, or noun phrases, and their syntactic functions and semantic roles are used to determine whether they are entities that carry core information. S140. Based on a pre-defined entity category thesaurus, perform semantic classification on candidate key entities to determine at least one key entity and its corresponding semantic category. Semantic categories include people, organizations, project names, and proprietary data. The entity category thesaurus contains a multi-level semantic tag system. The classification process is completed by string matching of entity names or semantic similarity calculation to ensure the accuracy of category classification.

[0018] Specifically, from the sentence "Project Leader C emphasized at last week's meeting...", identify the predicate verb emphasis and mark it as: Project Leader C: agent role; this is the highest priority matter: patient role; extract candidate key entities: R&D department, meeting, Project A, core architecture design document, key technology roadmap, strategic loss, marketing department, competitor B, person in charge C, last week's meeting and highest priority matter.

[0019] Based on the semantic classification of candidate key entities, the set of key entities is determined: e1: Project A, semantic category: project name; e2: Core architecture design document, semantic category: proprietary data (technical document); e3: Key technology roadmap, semantic category: proprietary data (technical information); e4: Strategic loss, semantic category: proprietary data (risk description); e5: Competitor B, semantic category: organization (competitor); e6: Person in charge C, semantic category: person (project leader); Specifically, step S200 includes: S210. Delineate a context window containing the key entity, centered on the location of the key entity in the email text; the size of the context window is adaptively adjusted according to the semantic density of the text and the entity type to capture the most relevant contextual information. S220. Match sentiment words and sentiment modifiers in the text within the context window to identify sentiment polarity words and sentiment intensity modifiers. The matching process is based on a preset sentiment dictionary and modifier dictionary, while also considering the influence of word negation modification and transition logic on the final sentiment tendency. S230. Based on the combination relationship between the identified sentiment polarity words and sentiment intensity modifier words, calculate the basic sentiment intensity value of the context window; the calculation process is based on the baseline score of the polarity words and the intensity coefficient of the modifier words to obtain a standardized sentiment intensity value. S240. Based on the semantic category of the key entity, assign an entity category weight factor to the basic sentiment intensity value to obtain the sentiment intensity coefficient of the key entity; the entity category weight factor is pre-set according to the sensitivity of the entity to sentiment information, reflecting the risk difference of different categories of entities; Specifically, taking the core architecture design document of entity e2 and strategic loss of e4 as examples, the context window of e4 is defined as: "This document...any disclosure could cause incalculable strategic losses"; the strong negative sentiment word "incalculable losses" is identified, and the basic sentiment intensity value is calculated to be -0.95, with negative values ​​indicating negativity; since category e4 is proprietary data (risk description), it is given a very high entity category weight factor of 2.0, resulting in a sentiment intensity coefficient E_s_e4=-0.95×2.0=-1.90; similarly, e2 also obtains a high sentiment intensity coefficient because it is modified by the highest priority.

[0020] Specifically, step S300 includes: S310. Perform time semantic analysis on the email content to identify and extract explicit time expressions and implicit time references; explicit time expressions include specific dates, times or period descriptions, and implicit time references include time information inferred through events, states or logical associations; S320. Establish the association mapping between temporal semantic information and each key entity. By analyzing the semantic connection relationship between time statements and key entities, determine the target time type associated with each key entity. The target time type includes past time, current time and future time. The semantic connection relationship is determined by analyzing the dependency distance, logical modification relationship and co-occurrence relationship between entities and time statements in the syntactic structure. S330. Based on the target time type and combined with the degree of association between time semantic information and key entities, determine the time sensitivity level of each key entity; the degree of association is determined by comprehensively quantifying the syntactic distance, logical dependency strength and semantic coherence between the entity and the time representation. S340. Map the time sensitivity level to a quantitative parameter, and perform weighted adjustment based on the time distance attribute of the target time type associated with the key entity to generate the time urgency coefficient of each key entity; the time distance attribute is calculated based on the difference between the target time type and the current system time, with future time having positive urgency and past time having negative or zero urgency.

[0021] Taking entity e2 as an example; the displayed time is identified as "before 18:00 on January 1, 2026", and a strong semantic association is established between it and e2, that is, the direct object of "completing final confirmation and submission" is "core architecture design document", and the target time type is future time; since it is a clear future deadline and closely related, the time sensitivity level is determined to be "extremely high"; the time distance attribute is calculated: if the current time is December 29, 2025, the deadline is about 3.5 days away; the formula for calculating the time urgency coefficient T_u is: T_u=α×exp(-β×Δt); where α is the base of the sensitivity level mapping, 1.0 for extremely high; β is the attenuation coefficient; Δt is the number of days away from the deadline; substituting into the calculation, we can get T_u≈1.0×exp(-0.5×3.5)≈0.85 for entity e2.

[0022] Specifically, the S400 includes: S410. Based on the semantic categories of key entities, map them to the corresponding semantic quantization values; S420. Based on the overall semantic context of the current email, dynamically allocate and adjust the dimensional weight factors of semantic quantification value, emotional intensity coefficient and time urgency coefficient; the overall semantic context is obtained through comprehensive analysis of the email's subject word distribution, stylistic features and writing intention, and the adjustment of the weight factors aims to reflect the differences in the importance of each dimension to the encryption decision under different contexts. S430. Using semantic quantification value, emotional intensity coefficient and time urgency coefficient as coordinates of the three-dimensional evaluation space, and using their respective corresponding dimension weight factors as coefficients, a three-dimensional evaluation matrix is ​​constructed. S440. Perform weighted fusion calculation on the parameters of each dimension in the three-dimensional evaluation matrix to obtain the comprehensive encryption sensitivity corresponding to each key entity; the weighted fusion calculation adopts linear weighted summation or nonlinear function mapping to ensure the reasonable integration of the contribution of each dimension; S450. If an email contains multiple key entities, the maximum value among all comprehensive encryption sensitivity scores shall be selected as the comprehensive encryption sensitivity score of the email.

[0023] Specifically, semantic categories are mapped to semantic quantification values ​​S_q; project name = 0.7, proprietary data (technology) = 0.9, proprietary data (risk) = 1.0; Analyzing the overall context of the email, it was determined to be an extremely high-risk context; dynamically allocating dimensional weights: semantic quantification weight W_s=0.4, emotional intensity weight W_e=0.4, and time urgency weight W_t=0.2. In this context, the weights of semantic and emotional factors are increased; Taking entity e2 as an example, the semantic quantification value S_q, sentiment intensity coefficient E_s, and time urgency coefficient T_u of each key entity are used to calculate the comprehensive encryption sensitivity C_s, representing the overall encryption requirement of the entity, through a weighted fusion model. The formula is: C_s = W_s × S_q + W_e × |E_s| + W_t × T_u; where C_s is the comprehensive encryption sensitivity, W_s, W_e, and W_t are the weights of the semantic, sentiment, and time dimensions, respectively, S_q is the semantic quantification value, |E_s| is the absolute value of the sentiment intensity coefficient, and T_u is the time urgency coefficient. The calculated comprehensive encryption sensitivity C_s of entity e2 is 1.25. After calculating all entities, it can be seen that the comprehensive encryption sensitivity C_s of entity e2 is the largest, so the comprehensive encryption sensitivity of the email is 1.25.

[0024] Specifically, the S500 includes: S510. Based on the combination relationship of semantic category, sentiment intensity coefficient and time urgency coefficient of key entities, a preset encryption rule base is constructed. The preset encryption rule base defines the encryption strength level and encryption validity period range corresponding to different combination relationships. S520. Based on the overall encryption sensitivity of the current email, and combined with the semantic category, sentiment intensity coefficient and time urgency coefficient of the key entities, match in the preset encryption rule base to locate the target rule entry. S530. Based on the target rule entries and with reference to the quantitative value of the sentiment intensity coefficient, determine the encryption strength level of the current email. The encryption strength level is positively correlated with the information security attribute of the semantic category and the quantitative value of the sentiment intensity coefficient. That is, the higher the inherent risk of the entity and the stronger the sentiment, the stronger the encryption algorithm or key length should be increased accordingly. S540. Based on the target rule entries and referring to the quantification results of the time urgency coefficient, determine the encryption validity period in the adaptive encryption strategy. The encryption validity period is negatively correlated with the quantification value of the time urgency coefficient. That is, the more urgent the time and the stronger the timeliness of the information, the shorter the encryption validity period should be set to achieve automatic expiration protection of information. S550 integrates encryption strength level and encryption validity period to dynamically generate an adaptive encryption strategy for the current email; the generated adaptive encryption strategy is a structured instruction set that specifies the encryption algorithm, key management method and access time window to be used for different paragraphs.

[0025] Specifically, the preset encryption rule base is a collection of logical judgments and parameter mappings. In this embodiment, the preset encryption rule base is defined as a two-level decision table: Level 1: Based on email comprehensive encryption sensitivity C_s mail Define the basic encryption framework; Rule R1: If C_smail If the value is ≥1.0, the highest strength encryption framework is triggered. The highest strength encryption framework requires the use of a hyperchaotic stream encryption algorithm based on semantic feature initialization to ensure the uniqueness and unpredictability of the key stream. Rule R2: If 0.5 ≤ C_s mail If the value is less than 1.0, the standard strength encryption framework is triggered; the standard strength encryption framework requires the use of an enhanced symmetric encryption algorithm. Rule R3: If C_s mail If the value is less than 0.5, the basic strength encryption framework is triggered; the basic strength encryption framework uses a standard symmetric encryption algorithm. Level 2: Within the established framework, determine specific parameters based on entity characteristics; Validity mapping function: Under the highest strength encryption framework, the base validity period T_base is set to an extremely short 8 hours; the dynamic adjustment formula for the final encrypted validity period T_expire is: T_expire=T_base×(1-T_u_w); where T_u_w is the weighted average of the time urgency coefficients T_u of all key entities in the email, ensuring that T_expire and T_u_w are negatively correlated; Encryption strength fine-tuning logic: The encryption strength level is fixed within the framework, but the main entity categories and sentiment strengths that lead to high sensitivity are recorded.

[0026] Furthermore, C_s is calculated based on the current email. mail =1.25, execute rule matching: match the first level rule: because 1.25≥1.0, rule R1 is the only match, that is, the highest strength encryption framework is determined; proceed to the second level parameter determination process of this framework; According to rule R1, the encryption strength level is directly determined to be the highest; corresponding to the explicit technical requirements: encryption algorithm: must use hyperchaotic stream encryption algorithm; key stream source: the initial state (x0, y0, z0, w0) and control parameters (a, b, c, d, r) of the chaotic system must be mapped from the dynamic seed S_dyn generated by mixing the semantic feature hash value of the email content with the spatiotemporal feature code; this strength level means that the encryption key stream is deeply bound to the email content, and the key streams are completely orthogonal if the content is different; The operation is performed according to the second-level mapping function under rule R1: calculate the overall time urgency coefficient of the email; perform a weighted average of entities e2 and e4 to obtain T_u_w=0.78; The final encryption validity period is calculated using the formula T_expire=T_base×(1-T_u_w)8×(1-0.78)=8×0.22=1.76 hours. Due to the high time urgency of the email content, the basic validity period is dynamically shortened from 8 hours to approximately 1.76 hours. This reflects the rule that the encryption validity period is negatively correlated with the time urgency coefficient, that is, the more urgent the information, the shorter the time window for access.

[0027] Specifically, the S600 includes: S610. Based on the semantic relevance of the email content and the distribution of key entities, the email text is divided into semantic paragraphs to ensure that key entities and their closely related semantic content are located in the same semantic paragraph. The division process is based on punctuation marks, paragraph marks and semantic transitions as boundaries, and is adjusted according to the referential relationship between entities and the coherence of the topic. S620. For each semantic segment after division, determine the corresponding segment encryption level according to the semantic category of the key entities contained therein and the encryption strength level determined in the adaptive encryption strategy. The segment encryption level is a refinement and adaptation of the overall encryption strength level of the email to the specific segment, and varies depending on the different entities within the segment. S630. Based on the paragraph encryption level, determine and execute the corresponding paragraph encryption processing method to achieve differentiated encryption for different semantic paragraphs; the encryption processing method includes selecting symmetric encryption algorithms, asymmetric encryption algorithms or hybrid encryption schemes of different strengths, and independently encapsulating the encrypted paragraphs; S640. Based on the encryption validity period determined in the adaptive encryption strategy, establish time-aware access permission rules; the access permission rules are implemented by binding the validity period information with the encryption key or access token through digital signature to ensure the immutability of time constraints. S650. Based on the access permission rules, access control is bound to the encrypted email content, so that the access permission is effective within the encryption validity period and automatically terminates after expiration, thus completing the binding of encryption and access control. After binding, when an access request is made, the current time will be verified to see if it is within the validity period, and the decryption operation will be allowed or denied accordingly, thus realizing fully automatic lifecycle management.

[0028] Specifically, taking the first paragraph containing e1, e2, and e3 (project and document description), and the second paragraph containing e4, e5, and e6 (risks and background information) as an example: According to the policy, both paragraphs are subject to the highest level of encryption. Semantic feature extraction and chaotic seed generation: Calculate the semantic fingerprint of the first text M1, such as using the secure hash algorithm SHA-256 to calculate H1=SHA256(M1); mix this hash value H1 with the spatiotemporal feature code of the email, and introduce a random salt value Random_Nonce to generate a dynamic seed S_dyn1; Chaotic system initialization: Map S_dyn1 to the initial state vector (x0, y0, z0, w0) and control parameters (a, b, c, d, r) of the selected hyperchaotic system; Since S_dyn1 is entirely determined by the specific content of the first segment, the starting point of the chaotic trajectory is generated for this segment; Adaptive chaotic encryption: The system begins iterating the chaotic equation to generate a pseudo-random sequence; for each plaintext byte P_i in the first segment, the chaotic state value is quantized into a key byte K_i; the ciphertext byte C_i = P_iXORK_i is calculated; An adaptive feedback mechanism is introduced: the current plaintext byte P_i is scaled by a small perturbation coefficient δ and fed back to the next iteration input of the chaotic system; for example, the state variable x_{i+1}=ChaosFunc(x_i,…)+δ×P_i is modified; so that the evolution of the key stream depends not only on the initial semantic seed, but also on the plaintext content being encrypted in real time, thus achieving memory and resistance to partial cracking. Segment encapsulation: The first segment of encrypted data is encapsulated independently, and its corresponding Random_Nonce and the initial parameter mapping information of the chaotic system are recorded; Based on the 1.76-hour validity period determined in the adaptive encryption strategy, a timestamp is bound to the decryption key of each segment, and a digital signature is generated to create time-aware access control rules. All encrypted paragraphs, the encrypted Random_Nonce, time binding rules, etc., are encapsulated into a final secure email. The recipient must decrypt the Random_Nonce and verify the validity of the time within 1.76 hours in order to recover the chaotic sequence using the correct initial parameters and then decrypt the email. After the expiration period, access control automatically expires.

[0029] This invention provides another technical solution: an adaptive encryption system based on the semantic features of email content. The system includes a feature extraction module, a semantic evaluation module, a policy generation module, and a segmented encryption module. The feature extraction module is used to receive emails to be encrypted, perform semantic role labeling on the email content, and extract at least one key entity and its semantic category. The semantic evaluation module is used to analyze the sentiment tendency of the text context in which each key entity is located to obtain the sentiment intensity coefficient, identify the time semantic information in the email content to determine the time sensitivity and generate the time urgency coefficient, and construct a three-dimensional evaluation matrix based on the semantic category, sentiment intensity coefficient and time urgency coefficient of the key entity to calculate the comprehensive encryption sensitivity. The strategy generation module is used to dynamically generate an adaptive encryption strategy that includes encryption strength and encryption validity period by matching preset encryption rules based on the comprehensive encryption sensitivity and the combination logic of entity category, sentiment coefficient and time coefficient. The segmented encryption module is used to divide email content into semantic segments based on an adaptive encryption strategy, perform differentiated encryption processing on semantic segments containing key entities, and bind time-aware access control.

[0030] Specifically, the semantic evaluation module includes a sentiment analysis unit, a time analysis unit, and a fusion evaluation unit. The sentiment analysis unit is used to define a context window centered on key entities, identify sentiment words within the window, and calculate the sentiment intensity coefficient. The time analysis unit is used to identify time semantic information in the email, establish its semantic association with key entities, and generate a time urgency coefficient based on time type and the degree of association. The fusion evaluation unit is used to quantify semantic categories into semantic quantification values, dynamically allocate the weights of each dimension according to the overall semantic context of the email, construct a three-dimensional evaluation matrix, perform weighted fusion calculation, and output the comprehensive encryption sensitivity.

[0031] Specifically, the segmented encryption module includes a segment analysis unit and a control binding unit. The segment analysis unit is used to divide semantic segments according to the semantic relevance of the email and the distribution of key entities, and to determine the segment encryption level according to the semantic category of key entities within the segment and the encryption strength level in the adaptive encryption strategy. The control binding unit is used to perform differentiated encryption according to the segment encryption level, and to establish and bind time-aware access control rules according to the encryption validity period in the adaptive encryption strategy.

[0032] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An adaptive encryption method based on semantic features of email content, characterized in that: The method includes: S100: Receive the email to be encrypted, perform semantic role labeling on the email content, and extract at least one key entity and its semantic category. S200. For each key entity, analyze the sentiment tendency of its text context and obtain the sentiment intensity coefficient of the key entity. S300: Identify the time semantic information in the email content, determine the time sensitivity of each key entity, and generate a time urgency coefficient; S400. Based on the semantic category of the key entity, the emotional intensity coefficient, and the time urgency coefficient, construct a three-dimensional evaluation matrix and calculate the comprehensive encryption sensitivity. S500. Based on the comprehensive encryption sensitivity, a preset encryption rule is matched based on the combination logic of entity category, sentiment coefficient and time coefficient, and an adaptive encryption strategy including encryption strength and encryption validity period is dynamically generated. S600. Based on the adaptive encryption strategy, the email content is divided into semantic segments, and differential encryption processing is performed on the semantic segments containing the key entities, and time-aware access control is bound to them. Step S600 includes: S610. Based on the semantic relevance of the email content and the distribution of key entities, the email text is divided into semantic paragraphs to ensure that key entities and their closely related semantic content are located in the same semantic paragraph. S620. For each semantic segment after division, determine the corresponding segment encryption level according to the semantic category of the key entities contained therein and the encryption strength level determined in the adaptive encryption strategy. S630. Based on the paragraph encryption level, determine and execute the corresponding paragraph encryption processing method to achieve differentiated encryption for different semantic paragraphs; S640. Establish time-aware access permission rules based on the encryption validity period determined in the adaptive encryption strategy. S650. Based on the access permission rules, bind access control to the encrypted email content so that the access permission is effective during the encryption validity period and automatically terminates after expiration, thus completing the binding of encryption and access control.

2. The adaptive encryption method based on semantic features of email content according to claim 1, characterized in that: Step S100 includes: S110. Perform syntactic analysis on the sentences in the email content and identify the predicate verbs in the sentences; S120. Taking the predicate verb as the center, analyze and label the semantic roles of other components in the sentence, including agent role, patient role, time, place and manner; S130. Based on the labeled semantic roles, extract nouns or noun phrases as core entities from the text components corresponding to the agent role and the patient role, and use them as candidate key entities. S140. Based on a preset entity category lexicon, perform semantic classification on the candidate key entities to determine at least one key entity and its corresponding semantic category. The semantic category includes people, organizations, project names, and proprietary data.

3. The adaptive encryption method based on semantic features of email content according to claim 1, characterized in that: Step S200 includes: S210. Delineate a context window containing the key entity, centered on the location where the key entity appears in the email text; S220. Match sentiment words and sentiment modifiers in the text within the context window to identify sentiment polarity words and sentiment intensity modifiers. S230. Calculate the basic sentiment intensity value of the context window based on the combination relationship between the identified sentiment polarity words and sentiment intensity modifier words; S240. Based on the semantic category of the key entity, assign an entity category weight factor to the basic sentiment intensity value to obtain the sentiment intensity coefficient of the key entity.

4. The adaptive encryption method based on semantic features of email content according to claim 1, characterized in that: Step S300 includes: S310. Perform time semantic analysis on the email content to identify and extract explicit time expressions and implicit time references; S320. Establish the association mapping between the time semantic information and each key entity. By analyzing the semantic connection relationship between the time representation and the key entities, determine the target time type associated with each key entity. The target time type includes past time, current time and future time. S330. Based on the target time type and the degree of association between the time semantic information and the key entities, determine the time sensitivity level of each key entity. S340. Map the time sensitivity level to a quantification parameter, and perform weighted adjustment based on the time distance attribute of the target time type associated with the key entity to generate the time urgency coefficient of each key entity.

5. The adaptive encryption method based on semantic features of email content according to claim 1, characterized in that: Step S400 includes: S410. Based on the semantic category of the key entity, map it to the corresponding semantic quantization value; S420. Based on the overall semantic context of the current email, dynamically allocate and adjust the dimensional weight factors of semantic quantification value, emotional intensity coefficient, and time urgency coefficient. S430. Using the semantic quantification value, emotional intensity coefficient, and time urgency coefficient as coordinates of the three-dimensional evaluation space, and using their respective corresponding dimension weight factors as coefficients, a three-dimensional evaluation matrix is ​​constructed. S440. Perform weighted fusion calculation on the parameters of each dimension in the three-dimensional evaluation matrix to obtain the comprehensive encryption sensitivity corresponding to each key entity; S450. If an email contains multiple key entities, the maximum value among all comprehensive encryption sensitivity scores shall be selected as the comprehensive encryption sensitivity score of the email.

6. The adaptive encryption method based on semantic features of email content according to claim 1, characterized in that: Step S500 includes: S510. Based on the combination relationship of semantic category, emotional intensity coefficient and time urgency coefficient of key entities, a preset encryption rule base is constructed. The preset encryption rule base defines the encryption strength level and encryption validity period range corresponding to different combination relationships. S520. Based on the comprehensive encryption sensitivity of the current email, and combined with the semantic category, sentiment intensity coefficient and time urgency coefficient of the key entity, match in the preset encryption rule base to locate the target rule entry. S530. Based on the target rule entries and referring to the quantized value of the sentiment intensity coefficient, determine the encryption strength level of the current email, wherein the encryption strength level is positively correlated with the information security attribute of the semantic category and the quantized value of the sentiment intensity coefficient. S540. Based on the target rule entries and referring to the quantization result of the time urgency coefficient, determine the encryption validity period in the adaptive encryption strategy, wherein the encryption validity period is negatively correlated with the quantization value of the time urgency coefficient. S550. Integrate the encryption strength level and the encryption validity period to dynamically generate the adaptive encryption strategy for the current email.

7. An adaptive encryption system based on semantic features of email content, characterized in that: The system includes a feature extraction module, a semantic evaluation module, a policy generation module, and a segmented encryption module; The feature extraction module is used to receive emails to be encrypted, perform semantic role labeling on the email content, and extract at least one key entity and its semantic category. The semantic evaluation module is used to analyze the sentiment tendency of the text context in which each key entity is located to obtain the sentiment intensity coefficient, identify the time semantic information in the email content to determine the time sensitivity and generate the time urgency coefficient, and construct a three-dimensional evaluation matrix based on the semantic category, sentiment intensity coefficient and time urgency coefficient of the key entity to calculate the comprehensive encryption sensitivity. The strategy generation module is used to dynamically generate an adaptive encryption strategy that includes encryption strength and encryption validity period based on the comprehensive encryption sensitivity and the combination logic of entity category, sentiment coefficient and time coefficient to match preset encryption rules. The segmented encryption module is used to divide the email content into semantic segments based on the adaptive encryption strategy, perform differentiated encryption processing on the semantic segments containing the key entities, and bind time-aware access control. The segmented encryption module includes a segment analysis unit and a control binding unit; The paragraph analysis unit is used to divide semantic paragraphs according to the semantic relevance of the email and the distribution of key entities, and to determine the paragraph encryption level according to the semantic category of key entities within the paragraph and the encryption strength level in the adaptive encryption strategy. The control binding unit is used to perform differentiated encryption based on the paragraph encryption level, and to establish and bind time-aware access control rules according to the encryption validity period in the adaptive encryption strategy.

8. An adaptive encryption system based on semantic features of email content according to claim 7, characterized in that: The semantic evaluation module includes a sentiment analysis unit, a time analysis unit, and a fusion evaluation unit; The sentiment analysis unit is used to define a context window centered on the key entity, identify sentiment words within the window, and calculate the sentiment intensity coefficient. The time analysis unit is used to identify time semantic information in emails, establish semantic associations between them and key entities, and generate a time urgency coefficient based on time type and degree of association. The fusion evaluation unit is used to quantify semantic categories into semantic quantification values, dynamically allocate the weights of each dimension according to the overall semantic context of the email, construct a three-dimensional evaluation matrix and perform weighted fusion calculations to output the comprehensive encryption sensitivity.

Images