Smart Grid is a power grid that improves flexibility, reliability, and efficiency through smart meters. Due to extensive data exchange over the Internet, the smart grid faces many security challenges that have led to data loss, data compromise, and high power consumption. Moreover, the lack of hardware protection and physical attacks reduce the overall performance of the smart grid network. We proposed the BLIDSE model (Blockchain-based secure quantum key distribution and Intrusion Detection System in Edge Enables Smart Grid Network) to address these issues. The proposed model includes five phases: The first phase is blockchain-based secure user authentication, where all smart meters are first registered in the blockchain, and then the blockchain generates a secret key. The blockchain verifies the user ID and the secret key during authentication matches the one authorized to access the network. The secret key is shared during transmission through secure quantum key distribution (SQKD). The second phase is the lightweight data encryption, for which we use a lightweight symmetric encryption algorithm, named Camellia. The third phase is the multi-constraint-based edge selection; the data are transmitted to the control center through the edge server, which is also authenticated by blockchain to enhance the security during the data transmission. We proposed a perfect matching algorithm for selecting the optimal edge. The fourth phase is a dual intrusion detection system which acts as a firewall used to drop irrelevant packets, and data packets are classified into normal, physical errors and attacks, which is done by Double Deep Q Network (DDQN). The last phase is optimal user privacy management. In this phase, smart meter updates and revocations are done, for which we proposed Forensic based Investigation Optimization (FBI), which improves the security of the smart grid network. The simulation is performed using network simulator NS3.26, which evaluates the performance in terms of computational complexity, accuracy, false detection, and false alarm rate. The proposed BLIDSE model effectively mitigates cyber-attacks, thereby contributing to improved security in the network.
In recent years, there is an increasing requirement of properly managing renewable resources and distribution of power based on the requirement. The smart grid has been evolved as a promising technology for proper management of the power resources by providing a decentralized supply of power from all sources. The smart grid network’s clean distribution and operational efficiency has been achieved by integrating the electric flow and communication network [
Few random tree methods were incorporated for attack detection. However, only known attacks were detected which will not be sufficient for the overall security of the smart grid network [
The prekeying technique emerged as an alternative keying method in which the key for the upcoming data was generated priory to reduce the latency involved in it; nevertheless, this method also had a key distribution problem [
Few existing works were used for the correction of parameters for the detection of most prevailing false data injection attacks. The integrity of the data was improved by executing the digital signature of packets for anomaly device detection in the smart grid-based edge network [
The goal of the research are;
To maximize the anonymity of the users by authenticating only the legitimate users in the smart grid network and to improve the security using blockchain technology, To maximize the integrity of the data by performing lightweight encryption of data without increasing the latency of the network, to overcome the key distribution problem, and to provide a secure transmission channel for transmission of data packets, To maximize the scalability in the smart grid network by optimal selection of the edge node (gateway) based on significant parameters, To maximize the security of the smart grid network in the control center side by implementing intrusion detection, To maximize the availability of the smart grid network in real-time by properly managing the users and providing updated and revocation phases.
To meet the abovementioned research objectives, in this paper, we proposed a Blockchain-based secure quantum key distribution and Intrusion Detection System in Edge Enabled Smart Grid Network (BLIDSE) model; we summarize our research contributions as;
Firstly, we proposed a lightweight block cipher algorithm for smart meters’ authentication, namely, Camellia cipher, which is in comparison with the Advanced Encryption Standard (AES) quite easier to use. It is a symmetric block cipher that provides better security compared to AES with an ultralow latency. Secondly, we particularly forwarded the data from the smart meters using an encrypted format and for particularly that, the procedure of Camellia is put to use. Confidentiality of the data is being secured and protected against transmission is ensured by implementation of lightweight Camellia encryption as well as the key distribution problem is overcome by performing quantum key distribution in which the key is transmitted in the form of photon, thereby providing unconditional security. Thirdly, we proposed a multi-constraints-based edge selection model which was presented for optimum selection to avoid any packet loss. The proper selection of edge nodes is carried out by performing a perfect matching algorithm that considers distance, load, makespan time, connectivity, delay and congestion, thereby facilitating reliable transmission of data. Here edge nodes are constructed into a hypergraph, and a perfect edge node is selected based on multiple constraints. Fourthly, we proposed Bi-Fold IDS in which the D-DQN is used in sync with historical data and the classification of data is done based on the dynamic threshold is performed, which will detect the new attacks as it computes the threshold for normal data. The D-DQN is trained with historical data from which the received data will be checked and the classification of data based on the dynamic threshold is performed, thereby the false alarm rate associated with the physical failure of devices will be reduced resulting in precise ID’s. The training of the proposed D-DQN is carried out with comparatively low time consumption, thereby improving the proposed work’s performance. The proposed work provides security to the users through blockchain-based secure authentication and provides security as well as the integrity of the data by executing Bi-Fold IDS which is performed to mitigate the physical attacks.
Finally, the optimal user privacy management ID’s carried out by FBI optimizer solves the optimization problem involved in updation and revocation to improve the security and scalability of the smart grid network. In the final analysis, the performance of the BLIDSE model is analyzed and evaluated during experiments, and we proved that as an outcome, it is more significant than the earlier models.
This paper is organized as follows; Section 1: Literature of intrusion detection in smart grid network, and edge assisted energy management is discussed, while in Section 2: A background of research problems is summarized and discussed in related work section. In Section 3: A detailed discussion of the proposed BLIDSE model is discussed, while, in Section 4: An experimental analysis of the proposed model and the previous works are compared. Further, this particular section covers the performance evaluation and security analysis of the proposed model. Finally, in the Section 5: there is a conclusion of the paper by stating the future directions and aspects of the same.
A multi-agent-based attack resilient system is presented for [
In [
A cyber-physical security model is presented in [
Two authorization methods are involved in validating the edge nodes in the proposed system, such as identity authorization and convert channel authorization. To manage security, and privacy, a blockchain-based anonymous authentication. With key management is presented for smart grid edge computing infrastructure. Mostly, attackers use hardware for capturing-sensitive energy-related information. A lightweight mutual authentication-based PUF scheme is presented for addressing the insider attacks. The unique properties of PUF are considered for authentication.
Machine learning (ML) approaches are used for intrusion detection and mitigation to protect cloud-based enterprise solutions [
The model-based data integrity method used public key encryption for the purpose of digital signature, but these methods are vulnerable to man-in-the-middle attacks, affecting the data integrity of the method. Also, it detects the attack by comparing the variation of statistics from the predicted statistics of the state; this causes unnecessary false alarms as the obtained values may vary due to other reasons. The lightweight authentication scheme as stated in [
It demands the need for user authentication in which only authorized users can access the smart grid network and provide anonymity for legitimate users, and thus the privacy of smart meters is protected. Secondly, the measurements of energy consumption and demand of every HAN are collected by the control center to generate and distribute the power from various sources. The data gathered from the HANs are encrypted before being transmitted in order to improve the confidentiality of the data.
Thirdly, edge servers
Through the BLIDSE model, three kinds of attacks are considered for detection and mitigation in this research
All the transactions are stored in the blocks with the format of hashing which is not tampered by any attackers. The hash code is generated and received by Double SHA-256 hash function. Each block is connected with existing blocks. Smart grid network faces many security challenges for that we introduced blockchain in smart grid network which improves security. Blockchain has many benefits described as follows; Blockchain is a distributed ledger that supports multifactor verification that improves security. It avoids deploying the third party into the verification process network, thus reducing energy and cost.
Blockchain has four metrics such as block validation processing time, storage cost, transaction processing time and rate of hash.
(a)
(b) Cost of Storage: This metric calculates the storage cost of the blockchain-based on data storage size. Storage cost for each block cycle is calculated as follows,
where,
(c) Processing time of Transaction: This metric evaluates the processing time of transactions, for that it calculates the time difference between the transaction starting time, and ending time which is defined as follows,
where,
(d) Rate of hash: Hashrate is calculated based on the count of block cycles which is defined as follows,
where,
Initially, the smart meter users
The control center collects the energy consumption and demand of every home area network (HAN) to generate and distribute the power from various sources. The data generated from the HAN are encrypted before being transmitted to improve the integrity of the data. For this purpose, a lightweight symmetric encryption cipher named Camellia is used, which is as secure as the standard AES, and computes with ultralow latency. The encryption is done by using the
This sequence of bits is used as key for encrypting the data. In this work, we consider the block length is 128 bits, and the key length is 128 bits. The encryption and decryption are done based on the secret key. The key is divided into two sub keys; each one has a 64-bit length. We consider two variables with the size of 128 bits, and four variables with the size of 64 bits, which is defined as follows,
This connection is determined for the length of the secret key k. For 128-bit length key consist
where, the four 128 bit length created the variables
where J represents the output bits, and I represent the input bits. The next function is the P function which is also run inside the F function which also takes the input (8 bit), and modifies the input as like the S function. In p function performs the XOR function, which is defined as follows,
Next function is the FL function which is performed during encryption and decryption, which is defined as follows,
Finally, the data is encrypted by using the camellia algorithm. The process of encryption and decryption of camellia is shown in
In the following, data encryption and decryption is discussed. After completing authentication, the user ( Step 1: Initially, Step 2: Blockchain server send a response to the user Step 3: Step 4: Server decrypts the Step 5: If is it matched then the blockchain server ready to share Step 6: During this process session key is generated by server that is used to encrypt the overall communication to provide confidentially of the data.
By doing so, the most prevailing attacks such as MITM, physical attacks, insider attacks, and eavesdropping are mitigated, hence the confidentiality of the data is ensured.
When multiple edge nodes are presented in the smart grid network, there is a huge likelihood for smart meters that may reside in the coverage area of multiple edge nodes. Some characteristics inherit from the proposed architecture as follows.
Smart meters transmit measurements from HAN to the nearby Measure data from sensors, actuators and other devices collects information by
The HAN network transmits the data to the perfect edge server from the available number of server based on the parameters such as
where
A mathematical formulation of the above bipartite matching problem is represented by follows.
where
In the control center, the received data are monitored to obtain each HAN’s energy consumption and further precede with billing and power generation processes. Therefore, the received data must be a credible one without any suspicious code in it. The false data injection attack is most commonly occurring in the smart grid network in which the smart meters are either compromised logically or physically, and the false data are manipulated with the original data before being transmitted to the control center. These types of attacks affect the performance of the power grid. To ensure that the received data is legitimate and doesn’t contains any false data. The Bi-Fold IDS is performed in which the packet flow based firewall is deployed which will drop all the irrelevant packets and this is performed based on the parameters such as
To ensure that the received data is legitimate and doesn’t contain any false data. The Bi-Fold IDS is performed in which the packet flow based firewall is deployed, which will drop all the irrelevant packets, and this is performed based on the parameters such as source IP, destination IP, IP protocol, source port number, destination port number, APDU type, ASDU type.
The next layer of IDS checks the integrity of the relevant message packets and classifies the data into three classes: normal, physical failure and attack. On both firewall features and packets are considered as
D-DQN belongs to the Q-learning family, and it is the updated method that works similar to the Q-target of DQN as follows,
where
For two different DQNs,
where
The algorithm for Bi-Fold IDS using Double DQN is given above. This algorithm is used to detect the abnormal packets transmitted from the smart meter is predicted. For temporal and spatial constraints between the packet, D-DQN learns the inputs and predicts the corresponding result. Through this process, the security of the smart grid network is ensured.
The availability is one of the important factors in the security; the scalability and availability of the smart grid network are increased by means through the optimal management of user privacy. This is carried out by performing updation and revocation of smart meters. The optimal updation and revocation is performed by implementing the Forensic Based Investigation optimization (FBI), which has improved convergence speed and convergence time than many other optimization algorithms. This algorithm has two phases such as investigation phase and the pursuit phase. The investigation phase investigates the suspected location. The investigation is executed if the user’s timeout for the key is attained or when the meter is suspicious.
In (L1), the new suspected location
where j = 1:d and d represent the dimension number, R and x represents the random integer in the range [−1, 1] and [0, 1], respectively.
Based on L1 and L2, the movement is defined as follows,
where,
where,
In (M1) every
where
In (M2), every agent
where,
The accuracy is an important factor in determining the intrusions of the smart grid system. The accuracy can be formulated as,
The experimentation of the proposed BLIDSE model is performed with extensive simulations. The simulation of the proposed BLIDSE model is carried out using NS 3.26. The validation of BLIDSE model is performed by creating a smart grid network of
The proposed BLIDSE model is found to have higher accuracy particularly; even when the number of compromised meters increases above 21 the accuracy is maintained with negligible reduction. This is due to the implementation of Bi-Fold IDS in which initially the firewall is deployed to filter the packets based on flow parameters, and further, the detection of intrusions is carried out by D-DQN packets are classified into three classes based on the dynamic threshold. This facilitates the accurate detection of both already occurred and new intrusions in the network. Once the meter is compromised, the FBI-based update and revocation are carried out to preserve the privacy of other users.
The accuracy of the proposed BLIDSE model is also compared with existing models for number of training samples, as shown in
The proposed BLIDSE model is found to have high accuracy mainly. When the number of training samples is 1000, the accuracy in detection of intrusion reaches to about 100%, whereas the existing models have accuracy lesser than the proposed model. This is due to the variation of threshold in detecting normal, physical failure and attack packets based on the training samples. Through this, the accuracy of identifying the attack packets is achieved, thereby improving the integrity of the user data.
The computational overhead is referred to as the additional load that restricts the reliability of the smart grid network. The overhead in the network is due to the increased number of requests from the meters in a particular time. This is caused mainly due to inappropriate selection of edge nodes, which causes interference resulting in increased latency in the transmission of data.
The attack packets in the smart grid network are detected in order to achieve confidentiality, integrity and availability. The number of detected attack packets is a measure of accurate detection of intrusions in the network. The number of detected attack packets increases with increase in the time taken for detection.
The false alarm rate is a significant metric in assessing the efficiency of a model. The false alarm rate in the smart grid network is caused due to the inaccurate classification of packets into attack packets which causes unwanted revocation of smart meters from the network.
The false detection of intrusions is affected by the threshold by which it is determined. The purpose of the threshold is to act as a boundary above which the packets are termed as attack packets. The threshold must be set accordingly to detect the attack packets accurately.
The comparison results of the proposed BLIDSE model in ensuring the security of the smart grid network is presented as a numerical representation in
Performance metrics | BLIDSE | BAKM | |
---|---|---|---|
Accuracy | Compromised meters | ||
Training samples | |||
Computational overhead (KB) | |||
Detected attack packets | |||
False alarm rate (%) | |||
False detection |
Due to the following reasons, the proposed BLIDSE model has obtained the better performance, and they are listed as follows,
The PUF based authentication of smart meters is carried out in which each node is authenticated in the blockchain, which mitigates insider attacks and physical attacks, thereby ensuring the anonymity of users. Previous works have presented centralized security mechanisms that do not resist security attackers. The key generation is carried out by implementing camellia cryptography in order to improve the confidentiality of the data, and the limitation of symmetric key distribution is overcome by implementing QKD, thereby mitigating MITM attacks. The user’s anonymity is improved by performing blockchain-based secure user authentication by considering the factors such as PUF, geographical location, and local time. The computational overhead of the smart grid network is reduced by optimally selecting the edge server, which is performed by using a perfectly matching theorem. The intrusion detection is executed by using Bi-Fold IDS in which the firewall is implemented to filter the packets based on flow parameters and the D-DQN is used to classify the packets into three classes namely normal, physical failure and attack based on the dynamic threshold thereby increasing the accuracy and F1 score. The updation and revocation of smart meters is executed by using FBI optimizer which ensures the data integrity and reduces false alarm rate in the smart grid network.
In this paper, the BLIDSE model is proposed for intrusion detection in edge enabled smart grid network. Our work solves the problems discussed in the above studies related to intrusion detection in smart grid networks. The proposed BLIDSE model mitigates MITM, physical, insider, and DDoS attacks in an effective manner, thereby contributing to improved security in the network. The main objective of this research is to provide security for smart grid networks. Blockchain-based secure user authentication is proposed to improve the network’s security. For this, we shared the secret key through the quantum channel during transmission. By using camellia, the user data is encrypted with the secret key, which improves the security and mitigates various attacks in the network such as MITM, insider attacks, and eavesdropping attacks. To achieve efficient transmission of data, we select the optimal edge server using a perfect matching algorithm. In this research, we perform two layers through Bi-Fold IDS, to improve the accuracy of the network. In the first level, the firewall is used to drop the irrelevant packets; in the second level of IDS, the integrity of the relevant message packets is checked, and the data is classified into three classes such as normal, physical error, and attack. For this we propose the DDQN algorithm, the D-DQN is used to classify the packets into three classes namely normal, physical failure and attack based on the dynamic threshold thereby increasing the accuracy and F1 score. Finally, user privacy is managed to improve security. The comparison results of the proposed BLIDSE model in ensuring the security of the smart grid network is presented as a numerical representation in
The authors sincerely acknowledge the support from Majmaah University, Saudi Arabia for this research.