Self-sovereign identity (SSI) introduces a novel identity management systems (IMS) paradigm, offering a privacy-preserving mechanism for identity verification [28]. SSI adheres to ten fundamental principles: existence, control, access, transparency, persistence, portability, interoperability, consent, minimization, and protection [29]. Existence denotes the independent status of users, and control refers to their ability to manage their identity. Access enables data retrieval, transparency ensures algorithmic and infrastructural clarity, and persistence guarantees long-term ID maintenance. Portability supports the transfer of identity-related information, while interoperability ensures broad usability. Consent signifies user agreement for identity use, minimization reduces data disclosure, and protection safeguards users’ rights.

Conventional authentication methods frequently employ sensitive personal identifiers encapsulated in encrypted tokens, which are vulnerable to information leakage if intercepted during transmission. Consequently, research on anonymous credential authentication—which verifies user eligibility without disclosing personal identities—has become increasingly prominent. Recent SSI techniques utilize anonymous credentials to protect personal information. These credentials enable users to authenticate themselves without unnecessary identity disclosure [28]. Generating anonymous credentials involves inputting the public key, message, proposition information, credential, and attribute proof signing key of the system. The attributes linked to the credential are identified based on the proposition details, and appropriate proof values are generated for each attribute. Upon receiving the anonymous credential, the server separates the attribute information according to the proposition details and conducts primary verification. Finally, the server authenticates the anonymous credentials by verifying the information of each attribute separately.

Fig. 1 illustrates the operation of SDRP through a flowchart. SDRP interacts with the authorization node (auth node) for identity verification, evaluates authentication permissions, and requests authentication from the user node. Initially, SDRP establishes rules to generate a digital ID that conceals personal information. Within this framework, a rule comprising all de-identification or specific identifiable operations related to the required attribute is stochastically generated and transmitted to the user node based on the authentication purpose. Certain de-identification rules permit the specification of attributes verifiable as sequences, such as age, gender, and affiliation. Upon receiving the rule, the user node computes a digital ID from its personal information and forwards it to the auth node. Subsequently, the auth node verifies whether the user falls within the authorized group range using the received digital ID. If the user node has an invalid ID outside the specified group range, the authentication process is flagged as abnormal, leading to authentication failure. Extending the auth request interval when the user node re-initiates authentication can prevent an infinite authentication loop caused by an abnormal node. Conversely, duplicate users are verified and authorized for authentication if the user node is identified as a legacy ID within the group range under verification. The duplicate user undergoes re-authentication considering the potential derivation of the same digital ID, despite the application of different rulesets. Unlike encrypting and transmitting personal identification information over the network, SDRP generates indirect personal information by performing operations on the attributes’ characteristics in the received rule set to derive a digital ID. Consequently, even if an attacker intercepts a digital ID through a side-channel attack, identifying an individual remains impossible, thus mitigating privacy infringement risks. Furthermore, SDRP achieves lightweight performance by employing simple arithmetic operations for digital ID generation.

Fig. 2 presents the system architecture of the auth and user nodes.

In this architecture, the auth node employs the ruleset generator to create rules that can be fully or partially identifiable, depending on specific authentication requirements. A fully identifiable rule is applied when attributes requiring authentication contain personally sensitive information that needs protection and belong to rule types where sequences undergo specific transformations. Conversely, partially identifiable rules are used for low-sensitivity attributes, posing no significant disclosure issues. This rule permits attributes such as age and gender to be directly identified through a digital ID sequence without additional operations. Fig. 3 provides an example of the SDRP digital ID generation process.

As illustrated in Fig. 3, rules are generated based on the required attributes to formulate an operation-based ruleset that ensures the anonymity of personal information. For instance, if the attribute is a string, a ruleset such as “Replace the spelling order of ‘name’ with numbers, then offset it at the sequence start” can function by substituting the alphabetical order with numbers. After the ruleset is created, it is transmitted to the user node via the sender. The user node applies the personal information stored in the personal information database (DB) to the ruleset within the digital ID generator to produce a digital ID, which is then transmitted to the auth node. The communication protocol between the auth node and the user node is depicted in Fig. 4.

Fig. 4 illustrates that the auth node manages authentication performance by incorporating the ruleset field into the communication packet format. Concurrently, the user node initiates authentication using the digital ID field. Upon receiving the digital ID from the user node, the discriminator assesses its validity and checks for duplicate IDs in the authentication history DB before making a decision.

Algorithm 1 outlines the pseudocode for SDRP encompassing rule generation, ruleset transmission to the user node, digital ID creation and return to the auth node, legitimacy assessment of the user node, and verification of duplicate digital IDs in the DB. In Step 1, fully or partially unidentifiable rules are generated and transmitted to the user node. In Step 2, upon receiving the ruleset, the user node applies a personal attribute to generate a digital ID, which is then returned to the auth node. In Step 3, the auth node calculates the ID condition to verify the authenticity of the user node. The calculated ID is compared with the digital ID received from the user node to determine authentication eligibility. In Step 4, if the ID is authorized for authentication, it undergoes a duplication check by comparing it with the digital ID stored in the authenticated history DB.

This section describes the experimental setup used to evaluate the performance of SDRP. The evaluation framework, depicted in Fig. 5, was implemented to compare and assess the performances of conventional and SDRP-based authentication methods.

The auth node of SDRP generates a ruleset for the required attributes and transmits it to the user node. The user node then computes the digital ID based on this ruleset, which the auth node subsequently uses to determine authentication permission. In this experiment, the parameter α, representing the number of required attributes, was varied to evaluate leakage risk, efficiency, and computational cost.

User data were obtained from the company–employee dataset [30], which includes information from 5000 users. Table 2 details the features of the data utilized in this experiment.

Personal information attributes of the user node, including ID, company, department, age, gender, and social number, were extracted from the total features. Only integer (int) and string (str) data types were employed. This dataset served as authentication data to simulate the authentication environment.

In this study, we benchmarked the zero-knowledge proof model (ZKPM) [31] and the identifiable attribute model (IAM) [27] to evaluate the performance of SDRP. ZKPM model employs an AES-based zero-knowledge proof technique for range proof. Here, the user node calculates y=gx(modp) to transmit y as an unidentifiable secret value x to the auth node. Subsequently, the user node generates a random number r and calculates C=gr(modp) to transmit C to the auth node. Upon receiving this, the auth node iterates the process N times, requesting either r or (x+r)(mod(p−1)) from the user node to estimate x and perform authentication. Parameters were set with g = 2, and N = 3. The random value r was chosen between 1 and 20, while p was selected from prime numbers between 100 and 200. User attributes from the company-employee dataset [30] served as the secret value x. Conversely, IAM [27] leverages lightweight zero-knowledge authentication protocols (ZKAP) with a session key for encrypting communication sessions, thereby reducing the number of authentication steps and ensuring confidentiality. IAM is designed for lightweight performance by dynamically adjusting the level of confidentiality. In this study, we compared the IAM environment at the lowest level of confidentiality.

To evaluate the performance of SDRP, we employed computational cost, leakage risk, and security efficiency as metrics. Computational complexity was assessed using Big-O notation.

Leakage risk denotes the potential for ID leakage by an attacker when utilizing SDRP, calculated as described in Eq. (1). (1)Leakagerisk=ItRIt×Ot(N)where RIt denotes the request interval at time t, Ot(N) represents the computational cost of the ruleset at time t, and It indicates the importance of the required attributes at time t. In this experiment, upon authentication failure, RIt increased by 10 s. Ot(N) is the computation cost based on Big-O notation, and It is determined by assigning importance to each attribute. Given that Social Number corresponds to sensitive personal information, its importance was set to 5, age and gender were set to 3, and company and department were set to 1 [32].

Security efficiency was calculated using Eq. (2) as a metric for evaluating the security efficiency of the authentication model. (2)Security Efficiency=PrivacypreservingcapabilityOt(N)

Security efficiency is inversely proportional to the latency and computational cost of authentication, and directly proportional to privacy-preserving capability, quantified as the proportion of preserved privacy, as calculated by Eq. (3). (3)Privacy preserving capability=1Leakage risk=RIt×Ot(N)It

SDRP presents an efficient authentication mechanism designed for lightweight devices to enhance security and minimize data leakage risk—a key security metric. The performance across three evaluation metrics was analyzed by progressively increasing the number of required attributes. To ensure experimental reliability, the average results from 10,000 simulation repetitions were calculated.

This section analyzes the performance of SDRP relative to conventional models, ZKPM and IAM, concerning computational cost, leakage risk, and security efficiency. Fig. 6 illustrates the comparative leakage risk between SDRP and conventional models.

As the number of required attributes increased, the probability of incorporating critical attributes also rose. Consequently, the risk of information leakage to potential attackers escalated, with the highest risk observed in the IAM model, followed by SDRP and ZKPM. The IAM model, which transmits and receives personal information at the lowest confidentiality level, exhibited the highest privacy leakage risk. In contrast, SDRP, which uses digital IDs containing only indirect personal information characteristics, and ZKPM, which relies on complex knowledge proof equations, demonstrated a maximum privacy leakage risk that was 102 times lower than that of IAM.

Fig. 7 illustrates the comparative results of SDRP and the conventional model regarding computational cost.

As the number of required attributes increased, the computational burden for de-identification escalated, consistently increasing overall computational costs. Among the methods, ZKPM, a zero-knowledge proof-based technique, exhibited inefficiencies due to the repeated exchange of complex operational formulas during secret-value de-identification. Conversely, SDRP, which computes a simple ruleset, reduced computational cost by up to 10 times compared to conventional ZKPM. Additionally, the IAM model, which bypasses de-identification processing, demonstrated optimal efficiency by avoiding separate operations. However, the most critical evaluation index for safe authentication, leakage risk, revealed a significant limitation: confidentiality cannot be guaranteed, as it presented the most inefficient results.

Fig. 8 illustrates the comparative security efficiency of SDRP vs. the conventional model.

Overall, an increase in the number of required attributes correspondingly elevated the computational cost of authentication, diminishing security efficiency. The SDRP model demonstrated up to five times higher security efficiency compared to conventional models. This enhanced efficiency is due to the superior leakage risk performance of SDRP relative to IAM, coupled with its lower computational cost compared to ZKPM. Conversely, while ZKPM achieved optimal leakage risk outcomes, its overall security performance was hindered by significantly higher computational costs.

In summarizing the experimental findings, this study assessed IAM, ZKPM, and the proposed SDRP method across three metrics: leakage risk, computational cost, and security efficiency. For leakage risk, the ranking was ZKPM, SDRP, and IAM, suggesting that more complex techniques offer stronger security. The evaluation of computational cost revealed the ranking as IAM, SDRP, and ZKPM, indicating a trade-off between security and efficiency. Therefore, security efficiency was the final metric, identifying the model that best balanced security and efficiency. In this metric, SDRP, ZKPM, and IAM were ranked accordingly. Since security efficiency is inversely proportional to authentication latency and computational cost while directly proportional to privacy preservation performance, it can be concluded that SDRP achieves the optimal balance between security and efficiency among the three models.