We use cookies to ensure our website works properly and to personalise your experience. Cookies policy
1. Rashtrasant Tukadoji Maharaj Nagpur University, Nagpur.
2. Professor, Nagpur.
The rapid propagation of Internet of Things (IoT) environments compels an increased emphasis on developing scalable and adaptable access control mechanisms to address heterogeneity and variability associated with device interactions. While traditional access control models operate efficiently they are unable to adapt to the changes in context and new security risks. On the contrary, blockchain-based approaches offer both decentralized trust and tamper-proof enforcement; however, these benefits come at the expense of considerable computational overhead and latency due to ongoing validation processes. This paper proposes a lightweight blockchain based predictive framework for context aware access control in IoT systems. This includes the use of a long short-term memory (LSTM) model to evaluate both temporal and contextual patterns within access requests. This allows for anomalous behavior to be identified prior to it occurring. A predictive filter is included in the architecture which filters out all access requests deemed irrelevant for blockchain based validation. This approach reduces unnecessary smart contract executions and improves overall system performance. To evaluate this proposed framework, a simulation environment was developed utilizing synthetic IoT access data. Results from this evaluation demonstrate that the proposed predictive framework provides a 28-32% reduction in authorization latency relative to blockchain only based access control architectures. Additionally, results indicate that throughput increases as load increases for the proposed predictive framework. Under controlled conditions, the predictive component demonstrates high accuracy in differentiating between normal and anomalous access patterns. These results indicate that by integrating predictive intelligence into decentralized trust mechanisms, an effective balance of security, scalability and computational efficiency may be achieved in IoT access control systems.
The rapid expansion of the Internet of Things has led to the proliferation of interconnected devices across smart cities, industrial systems, and intelligent infrastructures. These environments generate a continuous stream of access requests involving heterogeneous and resource-constrained devices operating under dynamic conditions. Ensuring secure, scalable, and adaptive access control in such settings has become a critical challenge.
Traditional access control mechanisms such as Role-Based Access Control (RBAC) and Attribute-Based Access Control (ABAC) are widely used due to their simplicity and low computational overhead. However, these models are inherently static and fail to adapt to rapidly changing contextual factors such as device behavior, location, time, and usage patterns. As a result, they are vulnerable to unauthorized access, replay attacks, and anomalous activities in large-scale IoT deployments.
To address the limitations of centralized and static access control, blockchain-based approaches have been introduced as a decentralized solution that provides tamper-resistant policy enforcement, transparency, and trust among distributed entities. Blockchain-based access control systems eliminate single points of failure and enhance auditability through immutable transaction records. However, these benefits come at the cost of increased latency and reduced throughput, as every access request typically requires smart contract execution and consensus validation.
Recent research efforts have attempted to combine context-aware access control with blockchain to improve adaptability. While these approaches incorporate environmental attributes into decision-making, they primarily rely on reactive mechanisms, where access decisions are made only after a request is received. This results to unnecessary processing overhead, especially in high-frequency IoT environments where a large number of requests may be benign or repetitive.
A major limitation of existing approaches is the absence of predictive intelligence for early identification of anomalous or high-risk access patterns. Without such capability, all requests are treated equally, resulting in inefficient utilization of blockchain-based resources and increased authorization delay.
To overcome these challenges, this paper proposes a lightweight blockchain-based predictive framework for context-aware access control in IoT systems. The proposed approach integrates a Long Short-Term Memory (LSTM) model to analyze temporal and contextual patterns in access requests and identify potential anomalies in advance. Based on this prediction, only relevant requests are forwarded for blockchain-based validation, thereby reducing unnecessary smart contract execution.
The primary contributions of this work are summarized as follows:
The remainder of this paper is organized as follows: Section 2 reviews related work, Section 3 presents the proposed methodology, Section 4 discusses experimental results, and Section 5 concludes the paper.
2. RELATED WORK
The problem of secure and scalable access control in Internet of Things environments has been widely studied, with existing approaches primarily categorized into traditional access control models, blockchain-based frameworks, and context-aware mechanisms.
Conventional models such as Role-Based Access Control (RBAC) and Attribute-Based Access Control (ABAC) have been extensively used in distributed systems due to their simplicity and low computational requirements [1], [2].
Although ABAC offers greater flexibility compared to RBAC, both models remain largely static and lack adaptability to dynamic IoT conditions. In large-scale environments where device behavior and contextual parameters change frequently, showing inefficiency capturing real-time variations, making them vulnerable to unauthorized access and anomalous activities.
These models are insufficient to capture real-time variations in IoT environments [3].
To address the limitations of centralized systems, blockchain-based approaches have been adopted to enable decentralized access control in IoT systems [5], [7]. Blockchain-based frameworks utilize smart contracts to enforce access policies, ensuring transparency, immutability, and tamper resistance. These systems enhance trust and eliminate single points of failure [8], [9].
Prior studies have shown that blockchain-based approaches can eliminate single points of failure and enhance trust among distributed IoT entities. However, these systems often suffer from high computational overhead and increased authorization latency, as each access request requires validation through consensus mechanisms and smart contract execution. This makes them less suitable for high-frequency IoT scenarios.
However, they introduce significant computational overhead and latency due to consensus mechanisms [10].
Context-aware access control models have been proposed to incorporate environmental attributes into decision-making [11], [12]. Recent research has focused on incorporating contextual information into access control decisions. Context-Aware Access Control (CAAC) extends traditional models by considering environmental factors such as location, time, and device status.
While these approaches improve adaptability, most of them rely on reactive decision-making. Access requests are evaluated only after they are generated, without anticipating potential anomalies or risks. As a result, unnecessary validation processes still occur, leading to inefficiencies in large-scale deployments.
However, these models remain reactive and lack predictive capabilities [13].
Despite significant advancements, existing solutions exhibit the following key limitations:
These limitations highlight the need for a framework that combines predictive intelligence with decentralized trust mechanisms to improve efficiency without compromising security.
Existing blockchain-based access control frameworks continue to suffer from scalability and performance issues [14], [15].
|
Model |
Key Features |
Strengths |
Limitations |
|
ABAC |
Attribute-based decision making |
Low latency, simple |
Static, no prediction |
|
Blockchain-based |
Smart contracts, decentralized validation |
High security, tamper-proof |
High latency, low throughput |
|
CAAC |
Context-aware decision making |
Adaptive to environment |
Reactive, no prediction |
|
Proposed Framework |
LSTM + Blockchain + filtering |
Balanced performance, reduced latency |
Requires training model |
Table 1. Comparative Analysis of Existing Access Control Models in IoT
Table 1 presents a comparative analysis of existing access control models based on key performance parameters, including scalability, adaptability, computational overhead, and security characteristics. Traditional models such as ABAC demonstrate low latency but lack dynamic adaptability. Blockchain-based approaches provide enhanced security and decentralization; however, they introduce significant computational overhead. Context-aware models improve flexibility but remain largely reactive. The analysis highlights the absence of predictive mechanisms in existing approaches, motivating the need for a lightweight predictive framework that balances security and performance
From the above discussion, it is evident that existing access control mechanisms either prioritize efficiency (as in traditional models) or security and decentralization (as in blockchain-based approaches), but fail to achieve an effective balance between the two.
Furthermore, the absence of predictive filtering mechanisms results in redundant blockchain interactions, increasing latency and reducing throughput.
To address these gaps, this paper proposes a lightweight predictive framework that integrates temporal learning with blockchain-based validation, enabling proactive decision-making and improved system performance in IoT environments.
METHODOLOGY
The proposed framework presents a lightweight and predictive access control mechanism for IoT environments by integrating contextual intelligence with decentralized validation. The architecture combines three key components: context-aware access control, predictive anomaly detection using a Long Short-Term Memory (LSTM) model, and blockchain-based policy enforcement.
The primary objective is to reduce unnecessary blockchain interactions by filtering access requests based on predicted risk levels. Instead of validating every request through smart contracts, the system evaluates contextual patterns and selectively forwards relevant requests for decentralized verification.
Fig.1 Proposed Predictive Access Control Framework for IoT System
Fig.1 illustrates the architecture of the proposed predictive access control framework for IoT systems. The framework consists of three primary layers: the IoT device layer, the predictive intelligence layer, and the blockchain validation layer. IoT devices generate access requests containing contextual attributes such as device ID, timestamp, location, and request frequency. These requests are analyzed by the LSTM-based predictive module, which analyzes temporal patterns to identify anomalous behavior. Based on the prediction, requests are categorized into normal and high-risk categories. Normal requests are processed through a lightweight authorization path without blockchain involvement, while high-risk requests are forwarded to the blockchain layer for secure validation using smart contracts and consensus mechanisms. The final access decision is recorded, and all validated transactions are stored in the blockchain ledger to ensure transparency and auditability.
The proposed system consists of three logical layers:
This layer includes heterogeneous IoT devices that generate access requests. Each request contains attributes such as device identity, timestamp, location, and access type. These attributes represent the contextual information required for decision-making.
This layer processes incoming requests and performs anomaly prediction using an LSTM model. The model analyzes temporal and contextual features such as request frequency and location patterns to identify abnormal behavior.
Requests identified as normal are processed with minimal delay, while anomalous or high-risk requests are flagged for further validation. This selective filtering reduces the number of transactions forwarded to the blockchain layer.
This layer enforces access control policies using smart contracts deployed on a consortium blockchain network. Only filtered requests are submitted for validation, ensuring reduced computational overhead. The blockchain provides tamper-resistant logging, decentralized trust, and secure policy enforcement.
The predictive component uses an LSTM-based model to capture temporal dependencies in access request patterns. Each request is represented as a feature vector including:
The LSTM model processes sequential input and produces a probability score indicating whether a request is normal or anomalous.
The risk associated with an access request is represented as:
Risk=fA_s, A_o,A_e
where A_s represents subject attributes (e.g., device ID), A_o represents object or resource attributes, and A_e represents environmental attributes such as location and time.
The classification decision is defined as:
y_t>θ→Anomalous Request
y_t≤θ→Normal Request
where y_t denotes the predicted probability score generated by the LSTM model at time t, and θ represents a predefined threshold used to distinguish between normal and anomalous access requests.
If the predicted probability exceeds a predefined threshold, the request is classified as anomalous and subjected to strict validation. Otherwise, it is processed through the lightweight path.
This mechanism enables proactive security by identifying potential threats before full validation, thereby reducing system load.
The end-to-end workflow of the system is as follows:
This workflow ensures that only necessary requests undergo expensive blockchain validation, improving overall efficiency.
The framework is designed to address the trade-off between performance and security in IoT access control systems.
The proposed approach balances these factors by introducing predictive filtering, which reduces redundant validation while maintaining decentralized trust.
RESULTS AND DISCUSSION
The proposed framework was evaluated using a simulated IoT environment with synthetically generated access request data. The dataset consists of 1,200 access requests, including both normal and anomalous patterns characterized by variations in request frequency, location, and temporal attributes.
To ensure compatibility with the predictive model, the dataset was preprocessed by encoding categorical attributes such as device ID, location, and access type into numerical form. Numerical features including timestamp and request frequency were normalized using Min-Max scaling to ensure uniform input representation.
The dataset was divided into training and testing sets using an 80:20 split, resulting in 960 training samples and 240 testing samples.
Three access control models were considered for comparative evaluation:
Traditional ABAC model, which performs rule-based authorization without predictive filtering
blockchain-only model, where all access requests are validated through smart contract execution and consensus
Proposed predictive framework, which integrates LSTM-based anomaly detection with selective blockchain validation
The performance of these models was evaluated using key metrics including authorization latency, system throughput, and prediction accuracy.
The predictive component of the framework was implemented using a Long Short-Term Memory (LSTM) model to classify access requests as normal or anomalous based on temporal and contextual features.
The model was trained for 10 epochs with a batch size of 32. During training, the model demonstrated stable convergence, with a consistent decrease in loss and improvement in classification accuracy across epochs.
The trained model achieved near-perfect classification accuracy under controlled synthetic conditions. This performance is primarily due to the clear separation between normal and anomalous patterns in the synthetic dataset, particularly in terms of request frequency and contextual attributes.
While the results validate the effectiveness of the predictive mechanism in a controlled environment, real-world IoT systems may exhibit more complex and overlapping patterns. Therefore, further validation using real-world datasets is necessary for comprehensive performance assessment.
Fig. 2. Latency Comparison of Access Control Models
The latency performance of the evaluated models was analyzed under varying request loads. As shown in Fig. 2, the blockchain-only model exhibits the highest authorization latency, with an average delay of approximately 120ms. This is primarily due to the requirement of executing smart contracts and performing consensus validation for every access request.
The traditional ABAC model demonstrates the lowest latency, averaging around 35ms, as access decisions are made locally without involving distributed validation mechanisms. However, this approach lacks adaptability and is less effective in handling dynamic and anomalous conditions.
The proposed predictive framework achieves an average latency of approximately 85ms, representing a reduction of nearly 28–32% compared to the blockchain-only model. This improvement is achieved through the predictive filtering mechanism, which minimizes unnecessary blockchain interactions by identifying and selectively processing access requests based on their risk level.
Fig. 3. Throughput Comparison under Varying Request Load
The throughput performance of the system was evaluated by measuring the number of requests processed per second under increasing load conditions. As illustrated in Fig. 3, the traditional ABAC model achieves the highest throughput due to minimal processing overhead.
In contrast, the blockchain-only model shows significantly lower throughput, as each request requires validation through consensus mechanisms and smart contract execution, limiting its scalability.
The proposed predictive framework demonstrates improved throughput compared to the blockchain-only model. By filtering non-critical requests before blockchain validation, the system reduces the workload on the distributed ledger, enabling more efficient processing of access requests under high-load conditions.
The experimental results highlight the inherent trade-off between performance and security in IoT access control systems. Traditional models such as ABAC provide low latency and high throughput but lack adaptability and robustness against dynamic threats. Blockchain-based approaches, while offering enhanced security and decentralized trust, introduce significant computational overhead that negatively impacts system performance.
The proposed framework addresses these limitations by integrating predictive intelligence with blockchain-based validation. The LSTM-based filtering mechanism enables proactive identification of anomalous access patterns, reducing unnecessary blockchain interactions and improving overall system efficiency.
The latency and throughput results further demonstrate that the proposed approach achieves a balanced performance by maintaining enhanced security than traditional models while significantly reducing the overhead associated with blockchain-based systems.
However, it is important to note that the current evaluation is based on a simulated environment using synthetic data. Real-world IoT deployments may involve additional challenges such as noisy data, device mobility, and network variability. Future work will focus on validating the framework using real-world datasets and exploring optimization strategies to further enhance scalability and efficiency.
CONCLUSION
This paper presented a lightweight blockchain-based predictive framework for context-aware access control in IoT systems. The proposed approach integrates LSTM-based anomaly prediction with decentralized Blockchain validation to address the limitations of traditional and blockchain-only access control mechanisms.
The experimental results demonstrate that traditional ABAC models achieve low latency and high throughput but lack adaptability in dynamic environments. In contrast, blockchain-based approaches provide enhanced security and decentralized trust however, they introduce significant computational overhead and increased authorization latency.
The proposed framework effectively balances these trade-offs by incorporating predictive filtering, which reduces unnecessary Blockchain interactions. The results indicate that the framework achieves a reduction in authorization latency of approximately 28–32% compared to blockchain-only models, while also improving throughput under increasing request loads.
The integration of predictive intelligence enables proactive identification of anomalous access patterns, improving system efficiency without compromising security. This makes the proposed framework suitable for large-scale IoT environments where both performance and security are critical.
However, the current study is based on a simulated environment using synthetic data. Real-world IoT systems may present additional challenges, including noisy data, dynamic device behavior, and network variability. Future work will focus on validating the framework using real-world datasets and improving its scalability under dynamic conditions.
REFERENCES
Srivalli Ch1*, Vinay Chavan2, A Lightweight Blockchain-Based Predictive Framework For Context-Aware Access Control In Iot Systems, Int. J. Sci. R. Tech., 2026, 3 (5), 277-284. https://doi.org/10.5281/zenodo.20044891
10.5281/zenodo.20044891