An AI-based blockchain framework with AES–RSA hybrid encryption for secure data sharing in smart environments

1 Introduction

The quantity of data generated, processed, and shared across connected devices has increased at an unprecedented rate due to the Internet of Things (IoT) and smart environments. Smart applications in healthcare and transportation facilitate intelligent decision-making, automation, and real-time service delivery. AES–RSA hybrid encryption for secrecy and safe key managementThese advancements enhance user experience and operational effectiveness, but they also pose serious challenges for data security, privacy protection, trust management, and system scalability. Smart environments are more susceptible to cyber risks, such as unauthorized access, data tampering, data leakage, and distributed denial-of-service (DDoS) attacks, because of the vast number of interconnected and heterogeneous devices, which increases the attack surface. Contemporary smart settings must now guarantee safe, dependable, and trustworthy data sharing. Traditional centralized security measures are sometimes ineffective because smart environments are vulnerable to attacks and single points of failure. Researchers have explored cutting-edge technologies like blockchain, artificial intelligence (AI), and cryptography to address these limitations. Blockchain provides decentralized trust, transparency, and data integrity, while AI enhances security through intelligent threat analysis and anomaly detection. Additionally, encryption methods are essential for safeguarding private data during transmission and storage. Although current research has achieved significant progress, these technologies are often investigated separately. Therefore, an integrated framework that combines decentralized integrity verification, secure encryption, and intelligent threat detection is still required. This article suggests an AI-driven blockchain infrastructure combined with AES–RSA hybrid encryption for secure data sharing in smart settings in order to solve these problems. AES–RSA hybrid encryption for data confidentiality and secure key management, blockchain technology for decentralized integrity verification and traceability, and AI-based file classification and anomaly detection are all included in this architecture.

The following summarizes the primary contributions of this work:

• Developing a cohesive AI-driven blockchain framework for secure and reliable data exchange in intelligent settings.
• Integrating AES–RSA hybrid encryption to enhance data privacy and ensure secure cryptographic key management.
• Applying AI-based methods for adaptive security, including anomaly detection and file classification.
• Ensuring decentralized integrity verification, transparency, and traceability through blockchain technology (Lilhore et al., 2026; Apak et al., 2026; Commey and Crosby, 2025; Bakas et al., 2026; Fu et al., 2026; Khan et al., 2024).

2 Literature review

Security challenges in smart environments and IoT systems have increased due to the rapid expansion of smart environments and the Internet of Things (IoT), However, because IoT devices are diverse and have limited resources, they present serious security and privacy issues. This has increased data interchange among linked devices. However, because IoT devices are diverse and have limited resources, they present serious security and privacy issues. Single points of failure and scalability issues plague traditional centralized security designs. Although current trust models are still susceptible to attacks and data manipulation, recent research has concentrated on decentralized and intelligent security methods. Consequently, there is a great need for safe and effective solutions that guarantee availability, confidentiality, integrity, and reliability in IoT contexts (AbdulKareem Al-abassy and Al-Jammas, 2026). Blockchain-based security in intelligent settings. Because of its decentralized, immutable, and cryptographic architecture, blockchain technology has become a promising solution to security and trust challenges in distributed systems. It has been extensively used in cloud, IoT, and smart applications to improve transparency, data integrity, authentication, and access control. Blockchain solutions still have to contend with issues including high processing costs in large-scale settings, latency, storage overhead, and scalability constraints. Furthermore, the integration of blockchain with intelligent security mechanisms has been prompted by its limited threat detection capability (Al-Abassy and Al-Jammas, 2024). Machine learning and artificial intelligence for cybersecurity. Because they make it possible to identify intricate and hitherto undiscovered dangers, artificial intelligence (AI) and machine learning (ML) are essential to contemporary cybersecurity. They are extensively utilized in network traffic analysis, anomaly detection, malware classification, and intrusion detection, especially in dynamic IoT environments. However, computing resources, feature selection, and data quality all affect how well they work. Furthermore, the majority of AI-based methods only handle threat detection; they do not address decentralized trust, safe storage, or integrity verification, which emphasizes the necessity of integrating with complementary security technologies (Shah et al., 2023). Hybrid encryption for secure data sharing. In smart environments, where sensitive information is frequently protected by cryptographic techniques, data secrecy is a basic necessity. RSA guarantees safe key exchange and authentication, while AES offers effective high-speed encryption. AES and RSA are used in hybrid encryption systems to provide robust security and high performance for data protection. Nevertheless, the majority of current methods do not integrate decentralized trust mechanisms and intelligent threat detection, underscoring the need for more complete security solutions (Kaushik et al., 2026). In order to combine decentralized security management with intelligent threat detection, recent research has concentrated on merging blockchain with artificial intelligence. In a variety of IoT areas, including smart cities, healthcare, and industrial systems, AI provides anomaly detection and predictive analytics, while blockchain guarantees trust, transparency, and unchangeable record-keeping. Despite the fact that this connection enhances security and trust management, many current frameworks continue to handle blockchain and AI independently with little to no cryptographic protection. Furthermore, there is still a lack of research on support for heterogeneous file processing in real-world settings (Wanjiru, 2025). Secure data exchange and file processing. In cloud computing and smart environments, where encryption and access control are frequently used to protect data, secure file sharing is essential. Nevertheless, the majority of current solutions only offer limited evaluation across various file kinds and sizes and concentrate on homogeneous datasets. Security frameworks that can effectively handle heterogeneous data are necessary due to the increasing diversity of digital material. Moreover, it is uncommon for current research to combine anomaly detection, secrecy, integrity, and authenticity into a single framework (Khayer et al., 2025). A blockchain-based smart fog-based video surveillance system with privacy protection. The study made use of blockchain technology to improve video transmission security and privacy protection in fog computing settings. Nevertheless, AI-based risk assessment, hybrid encryption, and secure processing of heterogeneous files are not included in the framework, which is primarily intended for surveillance data (Dave et al., 2022). presented a secure Internet of Medical Things (IoMT) platform for smart healthcare based on blockchain. Among linked healthcare equipment, the suggested framework enhances trust management, authentication, and medical data sharing. However, for improved data security and confidentiality, it is devoid of sophisticated encryption methods like AES–RSA hybrid encryption and intelligent anomaly detection (Dalal et al., 2023). Provided an effective blockchain-based framework for secure data transfer in Internet of Things settings. The strategy focuses on enhancing IoT device trust, data integrity, and communication security. However, it does not allow cloud storage integration, AI-driven file-level risk classification, or complete security of heterogeneous digital documents; instead, it focuses only on safe transmission (Bhattacharjee et al., 2024). In contrast to these current methods, the suggested framework combines cloud storage, blockchain-based immutable records, SHA-512 integrity verification, AES–RSA hybrid encryption, and AI-based risk classification to enable safe and intelligent data sharing in smart settings. Table 1 presents a comparison between the proposed framework and existing approaches.

TABLE 1. Comparison of the proposed framework with the available literature.

3 The problem statements

Large amounts of sensitive data are generated and exchanged in smart settings, making security, privacy, trust, and secure data sharing crucial issues. Current systems usually handle security needs independently, including hybrid encryption for data secrecy, blockchain for data integrity and trust management, or artificial intelligence for anomaly detection. However, the capacity to provide safe, scalable, and reliable data exchange is limited by the absence of a unified framework combining various technologies, especially when handling heterogeneous files in dispersed smart environments.

4 Research objectives

The purpose of this research is to improve the security of cloud-based data exchange by creating a secure blockchain framework powered by AI and combined with AES–RSA hybrid encryption. The study looks into how blockchain technology can be combined with cloud storage to offer traceability, confidentiality, integrity, and secure data management. The primary goals of this research are:

• To develop a safe framework for secure data sharing that combines blockchain technology, cloud storage, AES–RSA hybrid encryption, and artificial intelligence.
• To provide an AI-based risk classification and anomaly detection module that can classify uploaded data based on predetermined risk levels and identify possible security threats.
• To implement an AES–RSA hybrid encryption system to provide data privacy, safe key management, and effective sensitive information security.
• Utilize blockchain technology to create immutable transaction records that offer data integrity, authenticity, and traceability.
• To evaluate the suggested framework’s scalability, performance efficiency, security efficacy, and compatibility with other methods.

5 Blockchain technology

Blockchain technology allows for the safe, transparent, and unchangeable recording of transactions without the need for a central authority. It is a decentralized, distributed ledger. Data integrity, traceability, and resistance to manipulation are ensured by storing transactions in chronologically linked blocks and validating them through consensus methods among participating nodes. Blockchain, which was first introduced as the foundation of Bitcoin, has developed into a flexible platform that supports applications in supply chain management, cloud computing, IoT, and healthcare. Smart contracts, which are self-executing algorithms that autonomously enforce predetermined rules and agreements among users, are a fundamental component of blockchain technology. In decentralized settings, smart contracts improve security, lessen the need for middlemen, and enable trusted interactions. Additionally, blockchain offers dependable data-sharing capabilities, transparent audit trails, and secure identity management. Blockchain 1.0, which concentrated on cryptocurrencies, and Blockchain 2.0, which incorporated smart contract platforms like Ethereum, were the first two generations of the technology’s evolution. More recently, Hyperledger Fabric and other Blockchain 3.0 platforms have been created to enable permissioned, scalable, and flexible enterprise applications. Blockchain has garnered significant attention in cloud computing and IoT environments because of its decentralized architecture, which improves security, privacy, trust, and access control while lowering the danger of single points of failure. Because they offer better performance and efficiency than public blockchains, permissioned blockchain networks are especially appropriate for these kinds of settings (Amponsah et al., 2022). Blockchain is fundamentally a technical framework. A technology that lets users work together to maintain a decentralized, reliable database. Figure 1 illustrates how a typical blockchain system generates data. After which, they were grouped into storage blocks. Blocks are joined successively. Putting them in chronological order to create a linked data structure. Every user node participates in the assessment, archiving, and upkeep of data. Before a new block is produced, it usually needs to be approved. Broadcast to every user and over 50% of users, using nodes to synchronize throughout the whole network. Once synchronized, no changes or deletions are allowed (Shakor et al., 2024). Science Direct Blockchain Technical Framework Figure 1. Because distributed data storage, cryptography, and blockchain include technological breakthroughs from P2P networks, the consensus mechanism and appropriate processing power, storage, and bandwidth are sacrificed in favor of increased security. The following lists the main advantages blockchain has brought to security.

5.1 Blockchain technology characteristics (BCT)

Characteristics BCT possesses to manage a smooth and sustainable supply chain are described below and shown in Figure 2.

• Decentralization: Multiple systems can observe, track, save, and update the system’s data.
• Transparent: With the web’s approval, data are captured, saved, and retrieved on the networks. Additionally, it is always traceable and visible.
• Immutable: To confirm immutability, BCT includes time stamps and restrictions.
• Irreversible: An individual blockchain preserves a thorough and certifiable record of each transaction.
• Autonomy: Without a third party’s intervention, every blockchain node may independently acquire, send, save, and update the data.
• Open source: Every member of the network has open source access through blockchain while still keeping a hierarchy.
• Anonymity: The person’s identity is kept private when records are transferred among nodes.
• Ownership and uniqueness: Every shared document on the blockchain maintains a unique hash code and a record of its own transactions.
• Provenance: Every product on the blockchain has a digital record document attesting to its authenticity and provenance (Fahmi et al., 2023).

Figure 2 presents an overview of blockchain technology by highlighting its major qualities, including decentralization, transparency, immutability, autonomy, open-source architecture, anonymity, and data ownership. It also summarizes the main advantages of blockchain, such as data validation, improved response time, smart contract management, operational efficiency, immutability, and intellectual property protection, while identifying important limitations, including scalability challenges, high energy consumption, and interoperability issues among different blockchain platforms. Additionally, the image shows how blockchain can improve traceability and transparency at every stage of the supply chain—from the producer to the processor, distributor, retailer, and ultimately the consumer as well as contribute to social, environmental, and economic sustainability.

5.2 Blockchain-based IoT network security protocols

By offering decentralized identity management, access control, and data integrity, blockchain-enabled security protocols tackle major security issues in the Internet of Things. These protocols facilitate safe authentication, authorization, and tamper-resistant record-keeping via distributed ledgers and smart contracts. Research shows that in resource-constrained IoT scenarios, combining blockchain with edge computing and lightweight consensus mechanisms can improve security, trust, and traceability while lowering resource usage. Blockchain-based Performance parameters that directly impact scalability and responsiveness, like block generation time, transaction throughput, and confirmation latency, are frequently used to assess security protocols in Internet of Things contexts. The suggested AI-driven blockchain system for secure data sharing in smart environments is shown in Figure 3. An AI-based classification module initially evaluates data gathered from IoT devices, sensors, user files, and smart applications to identify abnormalities and evaluate security threats. AES–RSA hybrid encryption is then used to protect the classified data, guaranteeing both secure key management and data confidentiality. Blockchain technology uses smart contracts and immutable records to give decentralized access control, integrity, and traceability. Lastly, blockchain-verified access mechanisms are used to safely store and distribute the encrypted data to authorized users (Rajput et al., 2025; Chowdhury and Biswas, 2021).

6 Blockchain in AI

The combination of blockchain and artificial intelligence (AI) leverages AI’s data-processing capabilities with blockchain’s decentralized, secure, and tamper-resistant infrastructure. Blockchain guarantees data integrity, transparency, and provenance, in contrast to centralized AI, which is susceptible to data tampering and single points of failure. Secure decision-making and trusted automation are further made possible by smart contracts. In data-intensive environments like IoT and smart systems, this integration improves the dependability, security, and traceability of AI-driven applications. The development of AI approaches has accelerated due to the rapid rise of data created by sensor systems, IoT devices, social media, and web applications. AI includes both deep learning and machine learning. While DL automatically learns hierarchical features from raw data, enabling greater accuracy on challenging classification tasks, ML relies on manually engineered features. Because traditional AI architectures usually use centralized data processing and storage, they are susceptible to privacy issues, data manipulation, and single points of failure. On the other hand, a decentralized framework that guarantees data integrity, secure data sharing, transparency, and traceability is made possible by combining blockchain technology with AI. AI systems can use verified, immutable datasets via blockchain, promoting ethical AI, privacy protection, and trustworthy decision-making in distributed smart environments, as shown in Figure 4 (Khan and Murad, 2022).

7 Integration of blockchain and artificial intelligence to enhance security

One promising strategy for improving security, automation, and decision-making in smart settings is to combine blockchain technology and artificial intelligence (AI). While blockchain offers decentralized trust and transparency, AI techniques such as machine learning and deep learning enable intelligent data analysis, anomaly detection, risk assessment, and decision-making. In several fields, such as smart cities, healthcare, energy management, and IoT ecosystems, where blockchain ensures data integrity and traceability, and AI enhances system intelligence, recent research has demonstrated the efficacy of AI–blockchain integration. This collaboration enables the creation of safe, reliable, and scalable smart environments that can address emerging privacy and cybersecurity challenges. In addition to supplying SaaS, PaaS, and IaaS services, the cloud layer offers massive data centers and potent servers to support compute-intensive tasks, such as ML model training and Metaverse content storage. Through services like gaming, social networking, education, e-commerce, and blockchain-based apps, the application layer facilitates user interaction with the Metaverse. Access control, intrusion detection, identity authentication, and malware detection are examples of metaverse security techniques that safeguard users, devices, and resources. By providing tamper-resistant identification and data management, smart contract-based access control, and decentralized decision-making, blockchain improves security. Decentralized learning and intelligent security systems that enhance detection accuracy, privacy, and assault resistance are made possible by AI and blockchain. In addition to providing SaaS, PaaS, and IaaS services, the cloud layer offers massive data centers and potent servers to enable compute-intensive operations like training machine learning models. By facilitating decentralized decision-making, smart-contract-based access control, and tamper-resistant identification and data management, blockchain improves security. Decentralized learning and intelligent security systems that boost privacy, raise attack resistance, and improve detection accuracy are made possible by the combination of blockchain and AI. An integrated cybersecurity framework that incorporates Blockchain, Machine Learning (ML), and Decentralized Distributed Learning to improve the security, dependability, and privacy of contemporary networked systems is shown in Figure 5. The framework’s four main security features—identity authentication, intrusion detection, access control, and malware detection—address a variety of cyberthreats. Blockchain offers decentralized identity credential storage in the identity authentication layer, and machine learning (ML)-based models detect phony identities and impersonation attempts, reducing Sybil and impersonation assaults. The intrusion detection system uses ML-driven network traffic analysis and cooperative blockchain-enabled alert management to identify complex threats. Blockchain enables decentralized exchange of threat intelligence and malware signatures in the malware detection layer, while machine learning algorithms examine executable files, API call sequences, and byte-level characteristics to precisely identify malicious software, such as Trojan assaults and viruses. The framework’s fundamental component is the decentralized and distributed learning mechanism, which permits decentralized global aggregation and privacy-preserving local model training, protecting data confidentiality while enhancing the learning process’s scalability and resilience. The suggested architecture creates a reliable, intelligent, and robust cybersecurity ecosystem that can handle new threats in dispersed contexts by fusing blockchain with cutting-edge machine learning capabilities (Kuznetsov et al., 2024; Ebrahim et al., 2022; Truong and Le, 2024; Roman et al., 2018).

8 Risk detection module based on machine learning

For secure and scalable data sharing in intelligent settings, the proposed framework offers an AI-driven blockchain foundation combined with AES–RSA hybrid encryption. The system uses an adaptive pipeline that integrates blockchain-based integrity verification, cryptographic security, and machine learning-based classification to process heterogeneous files. Initially, input files are analyzed by a lightweight machine learning model, which assigns them to three risk levels (Low, Medium, and High) based on file attributes such as size and type. This classification allows for adaptive protection and establishes the necessary security level for each file. The file content is then effectively encrypted using AES, and robust confidentiality and secure key exchange are ensured by using RSA to encrypt and protect the AES keys. Lastly, a blockchain layer ensures data integrity and traceability by recording the file hash, timestamp, and categorization results in an unchangeable ledger. File processing in smart environments is secure, efficient, and scalable through the integration of blockchain, AI, and hybrid encryption. Prior to applying encryption and entering pertinent data on the blockchain, the module acts as an initial security assessment step. Because of its low computing complexity, efficiency in multi-class classification tasks, and capacity to handle diverse file-related information, the Random Forest classifier was chosen as the learning method. The ML model determines the proper risk rating for each file by using extracted file-level features, such as file size, file type, and file extension. This study used a controlled experimental setup to create a labeled dataset that replicates secure data-sharing scenarios in smart environments. The internally generated dataset consists of 10,000 file samples, including 4,000 Low-risk, 3,000 Medium-risk, and 3,000 High-risk samples. The dataset was tagged based on file characteristics and predetermined security criteria. Files were classified based on their size, category, and degree of sensitivity. Low-risk files included common formats and non-sensitive documents. Medium-risk files were those that contained executable formats with moderate security concerns or sensitive information. Files with several security-sensitive properties, executable files, encrypted archives, or extremely sensitive medical information were classified as High Risk. Before model training, these predetermined labeling guidelines were created to guarantee uniformity and repeatability. To maintain class distribution, stratified sampling was used to split the dataset into 80% training samples and 20% testing samples. The training subset was used to train the classifier, which was then assessed using testing samples that had not yet been viewed. The performance of the proposed classification module was assessed using standard evaluation criteria, including accuracy, precision, recall, and F1-score. The Random Forest classifier demonstrated effective performance in determining the risk level of uploaded files before applying encryption and blockchain-based security protocols, as shown in Table 2.

TABLE 2. A study of the random forest-based risk classification model’s performance.

The evaluation results are obtained using the independent testing subset. The Random Forest classifier demonstrates reliable multi-class risk classification performance across the three proposed security levels.

9 Results and discussion

Several input data samples were used to assess the suggested AI-driven blockchain infrastructure with hybrid encryption. Based on the length and complexity of the input data, the Artificial Intelligence module was able to classify the data’s security level and assign appropriate protection levels, such as Low, Medium, and High. Strong confidentiality was ensured by the efficient conversion of plaintext data into ciphertext using the AES encryption method. The encrypted output was unreadable, safely shielded from unwanted access. Improving system security overall as a result. Additionally, the blockchain module provided tamper-proof logging and guaranteed data integrity by recording a secure hash of the input data. The accuracy and dependability of the suggested framework were confirmed by the decrypted output, which precisely matched the original input data. Overall, the experimental results show that the suggested system accomplishes adaptive security management, integrity preservation, and secure data sharing in smart environments. The suggested framework shows its appropriateness for real-time smart environment applications by achieving an execution time of less than 0.01 s per operation. Even while integrating AI with Blockchain adds more computing steps, the overhead is still manageable. A lightweight private blockchain for safe record management, traceability, and data integrity. Only necessary metadata, such as file identifiers, timestamps, access details, and SHA-512 hash values, is stored on the blockchain; encrypted files are kept in cloud storage. To prevent unauthorized change, each block includes the timestamp, transaction details, current hash, and prior hash. To limit participation in transaction verification to approved entities, a permissioned validation mechanism is used. In order to assess the framework against common threats, such as replay attacks, Sybil attacks, man-in-the-middle attacks, key compromise, blockchain manipulation, double-spending, and smart contract vulnerabilities, a security analysis is carried out. The analysis shows that the combination of SHA-512 integrity verification, AES–RSA hybrid encryption, and blockchain immutability offers improved defense against these security risks. Man-in-the-Middle analysis, when an attacker eavesdrops on a legitimate party’s communication channel to obtain, alter, or insert unauthorized data, it is known as a Man-in-the-Middle (MITM) attack. The proposed system incorporates blockchain-based verification techniques and AES–RSA hybrid encryption to counter this attack. While AES offers effective encryption for the files being transported, RSA encryption is utilized to safely exchange the AES session key. Consequently, the encrypted content cannot be retrieved without the proper decryption key, even if an attacker manages to intercept the conversation. The likelihood of recovering the AES key during communication interception is analyzed to assess the security against MITM attacks. The size of the key space and the attacker’s computing power determine how much work is needed to crack the encryption. The attack model and its corresponding formulation are presented in Equation 1.

T y e a r s = T a t t a c k 60 × 60 × 24 × 365 (1)

9.1 System performance evaluation

The performance evaluation of the suggested AI-based blockchain security framework, which combines blockchain technology, AES–RSA hybrid encryption, and an AI-based anomaly detection module to enable secure data sharing, is summarized in Table 3. From 1 to 50 files have to be processed for the experimental evaluation. The framework demonstrated the computational efficiency of the hybrid encryption technique, with average encryption and decryption times of 1.68 m and 0.63 m, respectively. Data integrity may be verified quickly with no computing overhead, as demonstrated by the blockchain verification procedure, which took an average of 2.75 m. Additionally, with values ranging from 98% to 100%, the AI-based anomaly detection module achieved an average detection accuracy of 99%, demonstrating its efficacy in detecting unusual or malicious activity. The presented statistical metrics, such as the minimum, maximum, mean, standard deviation, 95% confidence interval, and p-value, show how the suggested framework is dependable and consistent throughout several experimental trials. The minimum, maximum, mean, standard deviation, 95% confidence interval, and p-value are among the statistical metrics that were computed to assess the dependability and consistency of the suggested framework over several experimental trials. Thirty separate replications of each experiment were conducted under the same experimental circumstances. The gathered measurements were used to calculate the mean and standard deviation, and the 95% confidence intervals were produced to show how uncertain the given values were. A two-tailed Student’s t-test with a significance level of α = 0.05 was used for the statistical significance analysis; p-values less than 0.05 were deemed statistically significant. Figure 6 shows the associated performance metrics.

TABLE 3. Statistical performance analysis of the proposed AI-based blockchain framework.

Thirty separate experimental runs yielded the statistical results. A two-tailed one-sample Student’s t-test with α = 0.05 was used to get the reported p-values. The suggested system exhibits consistent performance with minimal fluctuations in machine-learning classification, blockchain verification, encryption, and decryption.

9.2 File-level performance analysis

Table 4 provides a comprehensive file-level study of the suggested system, evaluating the computational performance and security categorization for various file sizes and kinds. Two similar but different labels, Status and Risk Level, are considered in this assessment. Each test file is assigned a predetermined security category, such as Normal, Suspicious, or Attack, which is represented by the Status label and serves as the evaluation reference class. The danger Level, on the other hand, is the comparable severity level produced by the AI-based detection module and is divided into three danger levels: Low, Medium, and High. The following is a definition of the mapping between these two labels: Attack files are classified as high risk, suspicious files as medium risk, and normal files as low risk. Each submitted file is given a suitable risk level by the AI module based on file attributes, such as file type and size. The findings indicate that while larger files, such as setup, take more processing time, smaller files, like data. csv, require less time (1.21 m encryption). Because of its larger size, exe takes a little bit more processing time (2.03 m to encrypt). Executable files (.exe) are categorized as high-risk since they belong to potentially dangerous file types. Additionally, the stability and effectiveness of the suggested framework are demonstrated by the encryption and decryption timings remaining constant across various file types. The link between file size, processing time, and the AI-based risk classification outcomes is shown in Figure 7.

TABLE 4. The proposed framework’s file-level security classification and computational performance.

AI, Predicted Risk Level is the result of the AI, classification module, whereas Security Status is the reference file category. Each file’s processing cost is shown by the encryption and decryption timings.

9.3 Performance evaluation and comparison

Despite the variety of file types, the times listed are mainly based on the overall size of the files and not their type, indicating that encryption and decryption algorithms handle data regardless of file type. Increasing size increases the time, the larger the size and number of files. A direct proportional relationship between file size and time implies that as the file size increases, the time required to process the file (e.g., upload, encryption, download, decryption) also increases proportionally. The processing time under various file sizes and kinds was examined to assess the scalability of the suggested framework. The findings show that even when the processed file size grows, the system’s encryption, decryption, and blockchain verification performance remain steady. Furthermore, the average execution time of individual security operations—a single encryption, decryption, or blockchain verification task—remained below 0.01 s. These findings show that the suggested framework can facilitate effective security processing for smart-environment applications in almost real-time, this relationship can be expressed mathematically as: T∝S. Where: where T is the encryption time, (ED) is the encryption data. The proposed encryption process is mathematically represented in Equation 2. The corresponding decryption process is formulated in Equation 3.

• T: Time required, S: Size file Time (T), measured during the encryption and decryption phases, is the amount of data processed per unit of time. The rate of encryption is calculated as:

T E = size ED T (2)

T D = size DD T (3)

Decryption throughput where size (DD)is the decryption data.

The computational overhead brought about by the blockchain verification procedure, cloud storage operations, hybrid AES–RSA encryption mechanism, and AI-based categorization module was used to assess the performance of the suggested framework. The sum of each execution stage is used to model the overall processing time of a medical document. The total processing time of the proposed framework is calculated by combining the AI, AES, RSA, blockchain, and cloud-storage processing times, as shown in Equation 4:

T tota = T AI + T AES + T RSA + T BC + T Cloud (4)

Table 5 evaluation of current works; the comparison reveals that traditional IoT security frameworks have poorer accuracy (65%) since they solely use encryption and lack intelligent detection capabilities. Although they increase detection accuracy to 99%, machine learning-based intrusion detection systems do not incorporate encryption. Although blockchain-based solutions have a high processing cost (3.50 m encryption time), they offer great data integrity. Although deep learning models have higher accuracy (85%), they frequently lack blockchain connectivity and come with a higher processing overhead. The results show that the suggested system achieves a balanced trade-off between security, efficiency, and scalability, making it appropriate for IoT and smart environments. The suggested model combines blockchain technology, lightweight machine learning, and hybrid AES–RSA encryption.

TABLE 5. Comparison of the proposed and existing security frameworks.

Recent blockchain-IoT and AI-based security studies were used to choose the baseline techniques. The numerical performance numbers are simply provided for reference because the original studies employed various datasets, experimental settings, and evaluation procedures. Therefore, rather than asserting a straight performance ranking, the comparison primarily concentrates on the security capabilities and architectural elements of each approach. Every experiment was carried out on a personal computer running Windows 11, an Intel Core i7 processor, and 8 GB of RAM. Python 3. x was utilized to implement the suggested framework. Standard cryptographic libraries for AES and RSA were used to carry out the cryptographic operations. To minimize measurement variability, each experiment was conducted several times, and the average execution time was recorded.

9.4 Ablation studies

An ablation study was conducted by systematically eliminating individual components while maintaining the remaining system configuration to examine the contributions of the major components of the proposed framework. The entire framework, which incorporates blockchain-based verification, SHA-512 integrity verification, AES–RSA hybrid encryption, and machine-learning-based risk assessment, was used as the reference configuration. The framework without AI-based risk classification, the framework without AES–RSA hybrid encryption, the framework without blockchain verification, and the framework without SHA-512 integrity verification were the four reduced configurations that were then assessed.

The Random Forest-based risk-classification module was eliminated in the first ablation setup, while the remaining elements were kept. The contribution of machine-learning-based risk assessment to the overall security architecture was evaluated using this setup. Classification accuracy was not applicable in this setup since the classification module was unavailable (N/A). The baseline encryption technique used in the experimental setup replaced the AES–RSA hybrid encryption process in the second configuration. This experiment assesses how hybrid encryption affects cryptographic processing efficiency and confidentiality. The third option keeps the remaining security features but eliminates the blockchain layer. This setup assesses how blockchain contributes to tamper-resistant record management, traceability, and integrity verification. To assess the contribution of cryptographic hashing to data-integrity protection, SHA-256-based integrity verification was used instead of SHA-512-based integrity verification in the fourth configuration to evaluate the hashing mechanism’s contribution to data integrity protection. For every ablation experiment, the same dataset, hardware configuration, software setup, and evaluation methods were used. To guarantee a fair and consistent comparison with the entire suggested framework, every configuration was assessed using the same experimental conditions. The same evaluation criteria were used to compare each configuration’s performance with the overall framework. These factors included blockchain verification time, encryption and decryption times, categorization performance, and, if relevant, overall processing behavior. Table 6 displays the ablation analysis results. By combining intelligent risk classification, hybrid cryptographic protection, cryptographic integrity verification, and blockchain-based traceability, the entire suggested architecture offers the most complete security configuration. While deleting the AES–RSA hybrid encryption process lowers the framework’s level of hybrid cryptographic protection, removing the AI classification component eliminates adaptive risk assessment. In a similar vein, switching from SHA-512 to SHA-256 modifies the chosen cryptographic integrity method, while removing the blockchain layer removes immutable auditability and blockchain-based traceability. Overall, the ablation results show that the suggested framework relies less on a single security technique and more on the complementary contributions of each of its constituent parts. Thus, the full configuration offers the most complete set of security, integrity, traceability, and intelligent risk-assessment features.

TABLE 6. AI-driven blockchain security framework component-based ablation analysis.

✕ denotes the removal of the appropriate component, whereas¹ implies its retention. N/A denotes that the setup does not apply the appropriate metric. The mean outcomes from the repeated experimental runs detailed in the statistical analysis section should be represented by the reported values.

9.5 Analysis of computational complexity

Four consecutive procedures are integrated by the proposed framework: cloud storage operations, blockchain verification, SHA, AES–RSA hybrid encryption, and AI-based risk classification. Big-O notation is used to examine each component’s computational complexity, as shown in Table 7. The overall computational complexity of the proposed framework is expressed in Equation 5, where:

• n represents the number of processed files.
• f represents the number of extracted features.
• m represents the file size.
• k represents the RSA key size.
• b represents the number of blockchain verification operations.

O n f + O m + O k 3 + O b (5)

TABLE 7. Computational complexity evaluation of the suggested security framework elements.

Note: The theoretical processing cost of each framework component is represented by the computational complexity. Here, n stands for the number of characteristics that were studied, f for the number of features that the AI classifier employed, m for the amount of the data, k for the size of the RSA key, and b for the blockchain verification processes. The suggested framework’s scalability and efficiency are examined using these complexity estimates.

Conclusion

This paper aims to solve the increasing difficulties of safe, reliable, and scalable data sharing in IoT-enabled smart environments. This research introduced an AI-driven blockchain framework integrated with AES–RSA hybrid encryption. The suggested framework integrates blockchain, cryptography, and artificial intelligence into a unified intelligent architecture, in contrast to traditional security methods that use these technologies separately. AES–RSA hybrid encryption offers effective data security and secure key management, while lightweight AI models carry out intelligent file classification and anomaly detection before adaptive encryption and blockchain registration. Blockchain technology also guarantees trusted data management, traceability, immutable auditing, and decentralized integrity verification. The results of the experimental evaluation showed that the suggested framework successfully balanced security, computational efficiency, and scalability, achieving an average encryption time of 1.68 m, an average decryption time of 0.63 m, an average blockchain verification time of 2.75 m, 99% anomaly detection accuracy, and 100% blockchain integrity verification. By combining intelligent threat detection, hybrid cryptographic protection, and decentralized trust inside a single architecture, the suggested framework surpasses traditional security solutions, according to additional comparative studies. All things considered, the suggested framework strengthens confidentiality, integrity, traceability, and trust in next-generation IoT ecosystems while offering a scalable and dependable basis for safe heterogeneous data sharing. Future work will focus on validating that the suggested framework offers near-real-time security processing based on the measured execution times and exhibits scalability by keeping consistent processing performance across various file sizes, lightweight blockchain consensus mechanisms, edge intelligence, federated learning, privacy-preserving techniques, and post-quantum cryptography to further enhance scalability, resilience, and real-time secure data sharing.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fbloc.2026.1935632/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Tianlin Li, Beihang University, China

Reviewed by: Nitin Goyal, Central University of Haryana, India

Ahod Alghuried, Prince Sattam Bin Abdulaziz University, Saudi Arabia