Sage Journals: Discover world-class research

Abstract

The existing privacy-preserving aggregation query processing methods in sensor networks rely on pre-established network topology and require all nodes in the network to participate in query processing. Maintaining the topology results in a large amount of energy overhead, and in many cases, the user is interested only in the aggregated query results of some areas in the network, and thus, the participation of the entire network node is not necessary. Aiming to solve this problem, this article proposes a spatial range aggregation query algorithm for a dynamic sensor network with privacy protection (energy-efficient privacy-preserving data aggregation). The algorithm does not rely on the pre-established topology but considers only the query area that the user is interested in, abandoning all nodes to participate in distributing the query messages while gathering the sensory data in the query range. To protect node data privacy, Shamir’s secret sharing technology is used to prevent internal attackers from stealing the sensitive data of the surrounding nodes. The analysis and experimental results show that the proposed algorithm outperforms the existing algorithms in terms of energy and privacy protection.

Keywords

Wireless sensor network query processing privacy protection spatial range aggregation query

Introduction

A wireless sensor network (WSN) is a multihop self-organizing network composed of multiple sensor nodes. An important feature of a WSN is data-centricity, and the nodes of the network cooperate with each other to store and process the sensing data of the monitoring area. Sensor networks are widely used in military defense,^1,2 intelligent transportation,³ medicine and health,⁴ and environmental monitoring.^5,6

In a WSN, users often submit a spatial range aggregation query to obtain statistical information of an area in the network, such as the average temperature and the maximum humidity of an area in a forest.

Sensor nodes are battery powered, having limited energy, and in many cases, their replacements of battery are difficult. Therefore, how to reduce the energy cost of the sensor network has always been a difficult problem; in the application scenario with information confidentiality requirements, how to improve the security of the sensor network information, prevent the sensor nodes from being captured and reduce the leakage of sensitive private data are additional problems. Therefore, it is necessary to study the privacy-preserving and energy-efficient spatial range aggregation query processing technique to solve these two problems.

The existing spatial range aggregation query algorithms Tiny AGgregation (TAG),⁷ R tree,⁸ and STWinFlood⁹ rely on preconfigured network topologies such as the tree rooted at the base station. But they do not take privacy protection into account. In addition, due to factors such as the environment and node failure, the network topology frequently changes. Maintaining a pre-established topology consumes considerable energy. In addition, the existing aggregation query privacy protection algorithm, Slice-Mix-AggRegaTe (SMART),¹⁰ and its variants privacy preserving, energy-efficient and scalable continuous data aggregation (PECDA)¹¹ and adaptive slice-based secure data aggregation (ASSDA),¹² hide original and sensing data by slicing the data. Although these algorithms have certain privacy protection capabilities, they all assume that the query area is the entire monitoring area, and all nodes in the network are required to participate in query processing. Obviously, these methods require considerable unnecessary energy overhead, and it is not necessary for all nodes to participate in query processing.

The existing privacy-preserving aggregation query processing methods rely on pre-established network topology and require all nodes in the network to participate in query processing. Maintaining the topology results in a large amount of energy overhead, and in many cases, the user is only interested in the aggregated query results of some areas in the network, and thus, the participation of the entire network node is not necessary. The main idea of the algorithm is as follows. The base station first sends a query message to a node in the query area, starting from the node and distributing the query message while gathering the sensing data in the query area. The algorithm uses the point-by-point overlay strategy to aggregate the data in the query area according to the established route and finally returns the final query result to the base station. The algorithm does not rely on the preconfigured topology, and the aggregation node sends the aggregation result to its next node instead of the original data, which ensures that the node’s sensing data will not be leaked.

The main contributions of this article are as follows:

We propose a route-based dynamic data collection protocol, which does not depend on the pre-established topology.

The privacy protection strategy proposed in this article does not require nodes to perform data slicing, confusion, and encryption, thus saving energy consumption.

The method proposed in this article only needs the nodes in the in query area to participate in query processing and does not require the nodes in the whole network, thus saving energy consumption.

The method proposed in this article provides efficient data aggregation while preserving data privacy.

The organization of this article is as follows. Section “Existing research works and their shortcomings” summarizes the related work, and section “Preliminaries” introduces the preliminaries. Section “Energy-efficient and privacy-preserving spatial range aggregation query processing in WSNs—E²PDA” proposes an energy-efficient and privacy-preserving spatial range aggregation query algorithm. Section “Analysis of experimental results” presents an analysis of the algorithm and its results, and section “Summary” summarizes the paper and presents future research directions.

Existing research works and their shortcomings

With the development of Internet of Things, IoT devices play an important role in smart cities. IoT devices in e-payment services are crucial for modernization.^13–15 Meanwhile, a large amount of data generated by these devices face the threats of privacy leakage.¹⁶ There are many works considering the protecting of sensing data. Due to the distrust between server and user,¹⁷ blind signature is crucial to protect user privacy. Blind signature can protect user’s anonymous instead of encrypting all the data and searching on the ciphertexts.^18–20 In addition, researchers proposed some methods to protect security in previous works,^21–24 which can provide us with new methods to make blind signature scheme in practice. Yun Wang et al. proposed a privacy-preserving framework for signature-based intrusion detection in a distributed network based on fog devices.²⁵ To prevent breaches of privacy, Yang et al.²⁶ proposed a novel blockchain-based privacy-preserving crowdsensing system. They use the anonymous nature of blockchains to protect the real identity of workers.

The objective of Data aggregation processes is to reduce the packet overhead and improve energy efficiency. Vodel and Hardt²⁷ discuss usual ways for data aggregation, including the adapted communication process. Guo et al.²⁸ use discrete particle swarm optimization (DPSO) to save node energy expenses and build an optimal routing tree assembly considering communication and aggregation cost.

Attacks against WSNs could be broadly considered from two different levels of views. One is the attack against the security mechanisms and another is against the basic mechanisms. Denial of Service (DoS^29,30) is produced by the unintentional failure of nodes or malicious action. Attacks proposed in Pfleeger and Pfleeger³¹ provide wrong information to the base stations or sinks.

The existing privacy-preserving aggregation query algorithms rely on pre-established topologies, which can be divided into two categories according to their topology type: tree-based algorithms and cluster-based algorithms.

Tree-based algorithms

He et al.¹⁰ proposed data aggregation privacy protection technology, SMART, based on data fragmentation. Each node divides its perceptual data into several fragments to hide its original sensing data and sends the data fragments to different intermediate nodes. After distributing the data fragmentations, the final aggregation result is finally derived at the base station. Considering that the data sent by the non-leaf node to its parent node is the result of the aggregation of the sub-tree where it is rooted rather than the original data, there is no need to protect the privacy of the data sent by the non-leaf node. Wang et al.¹¹ proposed a method for fragmenting the sensing data of the leaf nodes based on SMART called PECDA. The data fragments of the leaf nodes are sent to the neighbor nodes through the secure channel to protect the privacy of the leaf nodes. Ozdemir et al.³² proposed a data aggregation protocol, PRDA (polynomial regression-based privacy-preserving data aggregation), based on polynomial regression. Sensor nodes represented their data as polynomial functions to reduce data transmission, and the nodes sent coefficients of polynomial functions to intermediate aggregation nodes to avoid directly sending the sensing data of the nodes, thus protecting the privacy of the node data. Akila and Sheela³³ proposed the PDKP (preserving data and key privacy in data aggregation) protocol, which takes advantage of homomorphic encryption. The node’s data in the routing tree are encrypted for transmission, and the intermediate node gathers its children’s data and its own encrypted data until the final result is gathered at the sink node. Only the sink node can decrypt the encrypted data of each node. Hua et al.¹² proposed the ASSDA protocol, which dynamically determines the number of slices and the size of each piece of data, improving the performance of the SMART protocol.

Cluster-based algorithms

The CPDA (cluster-based private data aggregation) scheme in He et al.¹⁰ used the algebraic properties of the polynomial to calculate the required aggregate value. Ozdemir and Xiao³⁴ proposed the IPHCDA (integrity protecting hierarchical concealed data aggregation) protocol, which divides the network into multiple regions. The protocol allows hierarchical aggregation of encrypted sensor data while providing confidentiality and integrity. The data returned to the base station can be classified according to the encrypted key. The protocol uses homomorphic encryption to protect the sensing data privacy and authenticates the aggregated data by verifying the MAC to ensure the privacy of the data. LPDA (link-based privacy-preserving data aggregation) protocol was proposed in Zhang et al.³⁵ The sensing data of the nodes in the cluster subtract the base value that given by cluster head, then add a random number generated by the node, and then the data aggregation operation is performed. Elhoseny et al.³⁶ used the elliptical curve encryption algorithm to generate binary strings for each sensor, combined with a node ID, distance to the cluster head, and the index of the transmission to form a unique encryption key. Using XOR and other operations, encryption and decryption could be effectively implemented. Fang et al.³⁷ proposed a privacy-enhancing and energy-efficient data aggregation scheme, CSDA (cluster-based private data aggregation), which adopts privacy shard assembly technology, and the number of shards varies with the dynamic change of the network size, thus reducing communication and energy costs.

Many researchers have conducted very effective work and have laid a good foundation for subsequent research; however, their algorithms mainly rely on a pre-established tree or cluster topology, and the network topology frequently changes due to the influence of the environment, node failure, node movement, and other factors. Maintaining the topology incurs considerable energy overhead and reduces the lifetime of the network.

Preliminaries

To facilitate the discussion, we first present some basic concepts and preliminaries, including the attack model and the privacy protection protocol. In addition, this article mainly discusses the aggregation query and typical data aggregation functions including SUM, AVG, COUNT, and MAX/MIN. Without loss of generality, takes SUM as an example for discussion.³⁸

Attack model

Common attack modes in sensor networks include external attacks and internal attacks. External attacks refer to the private data of sensors obtained by attackers through eavesdropping. Because sensor networks use wireless communication, attackers acquire private data through eavesdropping on the data link layer when transmitting information between nodes. The eavesdropping attack is a common external attack mode in the sensor network. This article assumes that attackers can eavesdropping the entire network. An internal attack refers to an attacker acquiring information such as the sensing data and the key stored by the node through the capture node, which not only destroys the privacy of the captured node but also can infer the data flowing through the captured node by obtaining the secret key or colluding with multiple compromised nodes.

SMART protocol

He et al.¹⁰ proposed the privacy protection protocol SMART based on data slicing. SMART adopted the random secret key allocation strategy proposed in Eschenauer and Gligor,³⁹ which is mainly composed of three stages:

Secret key pre-allocation: There are K secret keys in the secret key pool, and each sensor node in the sensor network randomly selects k secret keys from the pool.

Discovery of shared secret keys: Each node finds a node in its neighbor that has the same secret key and establishes a secure connection if both nodes have the same secret key.

Establish the path secret key: Neighbor nodes without the same secret key can establish a secure connection through multi-hop routing.

Figure 1 shows the execution procedure of the SMART algorithm, the whole process of SMART is divided into three steps: data slicing, data mixing, and data aggregation. As shown in Figure 1, each node slices its own data into three pieces, keep a piece for itself, and then transmits the rest two pieces to its neighbor nodes. After all pieces are received, each node mixes its own piece and the received pieces to obtain a new result:

Data slicing: Each node randomly selects some nodes within its own adjacent or h-hop range, divides its own data into j parts, retains one piece of data, and distributes the remaining j – 1 pieces of data to those nodes within the h-hop range through secure channels.

Date mixing: After receiving the encrypted data slicing, the node uses the key of the sending node to decrypt it. The receiving node waits for a period of time to ensure that the remaining shard data arrive and then accumulate all received shard data.

Data aggregation: After the node’s data are mixed, the final query result is aggregated at the sink node according to the tree topology.

Figure 1.

The execution procedure of the SMART algorithm.

Based on SMART, Wang et al.¹¹ proposed an improved sensor network privacy protection protocol called PECDA. Figure 2 shows the execution procedure of the PECDA algorithm. There is no need to slice data for all the nodes in the query region. Only the leaf nodes slice their data in the data slicing stage. As shown in Figure 2, the leaf node n3 slices its own data into three pieces. d33 kept for itself, and then the rest two pieces are transmitted to n1 and n6.

Figure 2.

The execution procedure of the PECDA algorithm.

Both SMART and PECDA are privacy protection protocols based on topology trees. Because the corresponding topology will change due to node failure and movement, maintenance of the topology will incur additional overhead.

Energy-efficient and privacy-preserving spatial range aggregation query processing in WSNs—E²PDA

The idea of E²PDA

To avoid relying on a pre-constructed topology and to reduce the energy consumption of the nodes in query processing, this article proposes an energy-efficient and privacy-preserving spatial range aggregation query algorithm, named E²PDA.

Figure 3 shows the execution procedure of the E²PDA algorithm, assuming that the average temperature of the query region Q_a is calculated using E²PDA. The nodes in the query region are represented by a, b, and c, and the sensing data of the corresponding nodes is represented by D_a, D_b, and D_c. The algorithm divides the query region into several sub-regions, and the size of the sub-regions meets the condition that the distance between the adjacent sub-region nodes does not exceed the communication radius between the nodes. E²PDA does not rely on pre-established topologies. Its query route is also generated during query processing, which can effectively eliminate the influence of node movement and failure, and it uses Shamir’s secret sharing technology to prevent internal attackers from stealing the sensitive data of surrounding nodes and protect node data privacy.

Figure 3.

The execution procedure of the E²PDA algorithm.

The E²PDA algorithm is divided into three stages: (1) the query message’s geographic routing stage. The base station sends the query message to a node in the query region; (2) the query distribution and sensory data aggregation stage, the query message is sent to all nodes in the query region, and the sensing data is collected and aggregated to calculate the query results; (3) the query results generated in the second phase are returned to the query initiation node by geographic routing protocol.

Without loss of generality, the query area is described as a rectangle. The sink node of the initiating query node sends the query message to node a in the query region Q_a using the geographic routing protocol,⁴⁰ while the sink sends a random number r to node a to protect the data of node a. When the query message and random number arrive at node a, the random number is added to D_a to obtain D_a′. Then, node a sends the query message and D_a′ to node b. After receiving the query message and D_a′, node b adds D_b and D_a′ and sends it to the next node. The final query result in the query area is aggregated until E²PDA reaches the last node y in the query area. Node y sends the final aggregated result back to the base station via a geographic routing protocol. The sink node subtracts the random number r from the returned aggregation results to obtain the real aggregation results in the query region.

The entire dynamic query process of E²PDA does not depend on the pre-constructed topology, which aggregates the sensing data while ensuring the privacy of the node’s data. Taking node a as an example, the data sent by node a to node b is D_a′, and node b cannot infer the sensory data of node a. In addition, each node in E²PDA needs to send the sensing data only once. However, in SMART, each node needs to sends j – 1 copies (j is the number of copies of its own perceived data fragments) in the slicing stage, resulting in considerable energy overhead. Figure 4 shows a concrete diagram of the simulation route.

Figure 4.

A concrete diagram of the simulation route.

The design of E²PDA

This section describes the specific design of our algorithm, including four parts: the grid-based data collection strategy, the design of grid size, the method of handling voids and calculating the next node. The sensor network in the query region is represented as SN = {a, b, c,…}; we assume that the communication radius of the node is represented by R.

Grid-based data collection strategy

To be able to traverse the entire query area, a grid data collection strategy is proposed. The query area is divided into multiple grids, and the query message is sent to all nodes in the entire query area by means of the progression of the grid. Thus, the sensing data of the region of interest of the user are collected.

Grid size design

The width and height of the grid are denoted as gw and gh, respectively, and the adjacent grids are connected to each other, including the horizontal and vertical directions. After the query request arrives in the query region, all nodes in the query region need to send the query result and query request to the node in the next grid. The grid size meets the following conditions: $(2 gw)^{2} + (gh)^{2} \leq R^{2}$ and $(gw)^{2} + (2 gh)^{2} \leq R^{2}$ .

Method of handling voids

As shown in Figure 5, after the query node e collects the sensing data of the node in the grid EFGH, it needs to send the query message and partial query results to the next query node. As the GHKI under the grid EFGH does not have the neighbor node of node e, GHKI becomes a hollow region, which leads to the interruption of query processing. To avoid this problem, a geographic routing protocol is used to bypass the void region; that is, node e takes the center of rectangular region IKLM as the target location (where the distance between K and M is R), and the geographic routing protocol is used to send the query message and the partial query results to a node in IKLM. In Figure 5, node e arrives at node h in region IJLM through f and g and then continues the subsequent query processing from that node. If there is no node in the rectangular area IKLM or the geographic routing protocol cannot send the query message and part of the query result to the node in the area IKLM, a similar idea to the above method is adopted, and the center of the area LMNQ under IKLM is taken as the destination (where the distance between L and N is R), and the message is sent to a node in the area using a geographic routing protocol.

Figure 5.

Bypassing the void region.

Calculating the next node

If the query message and partial query results reach a node n. By querying the neighbor table maintained by node n, the neighbor node nearest to node n is selected as the next query node according to the route direction.

Energy consumption analysis

The energy of sensor nodes is mainly consumed by processing sensory data and transmitting data, and the energy consumed to transmit 1 bit of data is equivalent to the energy consumed to execute 800 computing instructions.⁴¹ Therefore, the greater the traffic, the greater the energy consumption. In the process of data aggregation, the energy consumption of data transmission takes up a large proportion. Therefore, we mainly focuses on the energy consumed in the transmission of data. Suppose that each node has an average of n_n neighbor nodes, S is the sink node, and there are N nodes in the query region. The energy consumed by the node for sending and receiving 1 byte is denoted as E_t and E_r, respectively. When broadcasting 1 byte of data, the sum of energy consumed by the broadcast node and the receiving node is E. It is easy to know that E = E_t + n_n × E_r. The size of query message is S_q, and the size of the sensory data message of each node in the query region is equal and is S_d.

The energy consumed by the SMART and PECDA schemes includes the establishment of secure channels for data transmission, the transmission of query messages, and the query results. Assume that the proportion of leaf nodes in SMART and PECDA is α. E_query represents the energy consumed in the query phase, and E_data represents the energy consumed in the data aggregation phase.

E(E²PDA), E(SMART), and E(PECDA) represent the total energy consumed by the three schemes. E_query(E²PDA), E_query(SMART), and E_query(PECDA) represent the energy consumed in the query distribution phases of the three schemes, and E_data(E²PDA), E_data(SMART), and E_data(PECDA) represent the energy consumed in the aggregation phases of the three schemes.

Theorem 1

$(E^{2} PDA) < E (PECDA) < E (SMART)$

Proof

E²PDA

E (E^{2} PDA) = E_{query} + E_{data}

(1)

where

E_{query} = S_{q} \times E \times N

(2)

E_{data} = N \times \bar{s_{a}} \times E

(3)

SMART

E (SMART) = E_{query} + E_{data}

(4)

where

E_{query} = S_{q} \times E \times N

(5)

E_{data} = N \times \bar{s_{a}} \times E \times J

(6)

PECDA

E (PECDA) = E_{query} + E_{data}

(7)

where

E_{query} = S_{q} \times E \times N

(8)

E_{data} = N \times ((1 - α) \times \bar{s_{a}} + α \times \bar{s_{a}} \times j) \times E

(9)

From equations (2), (5), and (8), it can be deduced that

E_{query} (E^{2} PDA) = E_{query} (PECDA) = E_{query} (SMART)

(10)

From equations (3), (6), and (9), it can be deduced that

E_{data} (E^{2} PDA) < E_{data} (PECDA) < E_{data} (SMART)

(11)

The conclusion can be drawn from equations (10) and (11)

E (E^{2} PDA) < E (PECDA) < E (SMART)

Privacy analysis

The literature³⁹ contributes two important properties of key distribution: the probability of at least one identical key between any two neighbors is $P_{Connect}$ , and the probability that the third party has the same key is $P_{overhear}$ , detailed as follows

P_{Connect} = 1 - \frac{{((K - k)!)}^{2}}{(K - 2 k)! 2 K!}

P_{overhear} = \frac{k}{K}

In the SMART algorithm, when establishing a secure connection, node S_i and node S_j can establish a secure connection because S_i and S_j have the same secret key value d_ij. The privacy leakage of node data mainly includes two aspects: (1) the third-party node has the same secret key with the probability $P_{overhear}$ ; (2) because of the secret key preallocation stage, each node randomly selects k secret keys from the secret key pool with K secret keys and then establishes a connection if the adjacent nodes have the same secret keys. The third party can guess the secret key, and the correct guess is $1 / C_{K}^{k}$ . Therefore, SMART, and PECDA have the possibility of privacy data leakage during data aggregation. In E²PDA, if an attacker wants to get the original data of a node n, he needs to get the aggregation result sent to the node n and the data transmitted to n’s next node after node n completes the aggregation operation. However, E²PDA adopts a route-based dynamic aggregation method, which increases the difficulty of the attacker guessing the next routing node and avoids the problem of sensing data leakage caused by secret key leakage in SMART and PECDA. In terms of the security of private data, E²PDA is superior than SMART and PECDA.

Analysis of experimental results

In this section, the energy consumption of the E²PDA is analyzed through experiments. E²PDA, SMART, and PECDA are implemented in the simulator.⁴² The experiments are conducted on a PC with a P4 2.6 GHz CPU and 8 GB memory running Ubuntu operating system. The sensor nodes are deployed in a rectangular monitoring area in a uniformly distributed and non-uniform distribution, respectively. And the experiment uses dataset provided by the University of Southern California ANRG Laboratory.

The experimental parameters were selected as follows. According to Rappaport,⁴³ the energy consumption formula of wireless communication transmission and reception of 1 byte is $E_{t} = α + γ \times d^{n}$ , $E_{r} = β$ . Where $α$ and $β$ capture the energy dissipated by the communication electronics, $γ$ represents the energy radiated by the power-amp, n is the path loss index, and d is the distance to which a bit is being transmitted. The parameters in Coman et al.⁴² were adopted: $γ = 10 pJ / bit / m^{2}$ , $α = 45 nJ / bit$ , $β = 135 nJ / bit$ , $n = 2$ . The other parameters are shown in Table 1.

Table 1.

Experimental parameters.

Parameter name	Parameter value
Network coverage area	450 m × 450 m
Network node distribution	Random uniform distribution
Node communication radius	50 m
Number of nodes	600
Perceived data size	30
Query data size	24

Figure 6 shows the effect of the number of network nodes on energy loss. When the number of network nodes is fixed, PECDA and SMART consume significantly more energy than E²PDA. There are a certain number of nodes. Under the same query area, the SMART algorithm needs to distribute the slicing data to the neighboring nodes in the data slicing stage, and the energy consumed in this process accounts for a large proportion of the total energy consumption. The PECDA algorithm conducts slicing transmission only of leaf node perception data, while the SMART algorithm requires data slicing transmission of all nodes in the network.

Figure 6.

Influence of the number of network nodes on energy loss: (a) uniformly distributed network and (b) non-uniformly distributed network.

Figure 7 shows the energy consumption of E²PDA, SMART, and PECDA in different network topologies. As shown in the figure, the change in network topology has little effect on the energy loss of E²PDA. However, SMART and PECDA are susceptible to the network topology due to their dependence on pre-constructed topology trees. Different topology structures have different energy losses. In addition, according to the figure, E²PDA performs better than SMART and PECDA in energy consumption when the topology is certain.

Figure 7.

Influence of different network topologies on energy loss: (a) uniformly distributed network and (b) non-uniformly distributed network.

Figure 8 shows the energy consumption performance of E²PDA, SMART, and PECDA in the case of query messages of different sizes. According to the figure, the energy loss of SMART and PECDA is greater than that of E²PDA. The energy consumed by the three algorithms in the process of sending a message is the same, and the energy consumption difference between the three algorithms is mainly derived from data slicing and distribution in SMART and PECDA.

Figure 8.

Impact of the query message size on energy loss: (a) uniformly distributed network and (b) non-uniformly distributed network.

As shown in Figure 9, when the perceived data size is certain, the energy consumption of E²PDA is lower than that of SMART and PECDA. When the sensing data stored by the node increases, the energy consumption of SMART and PECDA in the data slicing and distribution phase occupies a certain proportion in the total energy consumption because SMART and PECDA include three stages: slicing, mixing, and aggregation. As the sensing data increase, the energy consumption of the data slicing and distribution also increases. Compared with SMART and PECDA, there is no need to divide and distribute the node-aware data during E²PDA data aggregation, which decreases the energy overhead.

Figure 9.

Influence of the perceived data size on energy loss: (a) uniformly distributed network and (b) non-uniformly distributed network.

As shown in Figure 10, the energy consumption of the three algorithms increases with the increase in the query region, and when the query region is constant, the energy consumption of E²PDA is lower than that of SMART and PECDA. In a uniformly distributed sensor network, the query area becomes larger, the number of nodes to be aggregated increases, and the energy consumption increases as the number of nodes in the query area increases.

Figure 10.

Influence of the query area size on energy loss: (a) uniformly distributed network and (b) non-uniformly distributed network.

Summary

We proposed an energy-efficient and privacy-preserving spatial range aggregation query processing algorithm in WSNs. It does not depend on the pre-constructed topology structure and collects the sensing data of the specified area while ensuring the data privacy of the sensor nodes. In the query area, if attacker wants to get the original data of a node n, he needs to get the aggregation result sent to the node n and the data transmitted to the n’s next node after node n completes the aggregation operation. Dynamic route-based aggregation method increases the difficulty of guessing the sensing data. The analysis of the algorithm and experimental results shows that our algorithm outperforms some existing algorithms such as SMART and PECDA in terms of energy consumption and privacy protection.

Footnotes

Handling Editor: Weizhi Meng

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research work was supported by the Aeronautical Science Foundation of China (grant no. 20165515001), the National Natural Science Foundation of China (grant no. 61402225), State Key Laboratory for smart grid protection and operation control Foundation, Science and Technology Funds from National State Grid Ltd. (The Research on Key Technologies of Distributed Parallel Database Storage and Processing based on Big Data).

ORCID iD

Liang Liu

References

Butun

Morgera

Sankar

. A survey of intrusion detection systems in wireless sensor networks. IEEE Commun Surv Tut 2014; 16(1): 266–282.

Akyildiz

Sankarasubramaniam

et al . Wireless sensor networks: a survey. Comput Netw 2002; 38(4): 393–422.

Lee

Zhou

Gerla

et al . Mobeyes: smart mobs for urban monitoring with a vehicular sensor network. IEEE Wirel Commun 2006; 13(5): 52–57.

Stankovic

Cao

Doan

et al . Wireless sensor networks for in-home healthcare: potential and challenges, 2005, pp.2–3, https://www.researchgate.net/publication/228619847_Wireless_sensor_networks_for_in-home_healthcare_Potential_and_challenges

Szewczyk

Mainwaring

Polastre

et al . An analysis of a large scale habitat monitoring application. In: Proceedings of the 2nd international conference on embedded networked sensor systems, Baltimore, MD, 3–5 November 2004, pp.214–226. New York: ACM.

Sullivan

Wood

Iliff

et al . eBird: a citizen-based bird observation network in the biological sciences. Biol Conserv 2009; 142(10): 2282–2292.

Madden

Franklin

Hellerstein

et al . TAG: a tiny aggregation service for ad-hoc sensor networks. ACM SIGOPS Oper Syst Rev 2002; 36(SI): 131–146.

Goldin

Song

Kutlu

et al . Georouting and delta-gathering: efficient data propagation techniques for geosensor networks. Boca Raton, FL: GeoSensor Networks, 2005, pp.73–95.

Coman

Nascimento

Sander

. A framework for spatio-temporal query processing over wireless sensor networks. In: Proceeedings of the 1st international workshop on data management for sensor networks: in conjunction with VLDB 2004, Toronto, ON, Canada, 30 August 2004, pp.104–110. New York: ACM.

10.

Liu

Nguyen

et al . PDA: privacy-preserving data aggregation in wireless sensor networks. In: Proceedings of the INFOCOM 2007–26th IEEE international conference on computer communications, Barcelona, 6–12 May 2007, pp.2045–2053. New York: IEEE.

11.

Wang

Qin

Ding

et al . Privacy-preserving and energy-efficient continuous data aggregation algorithm in wireless sensor networks. Wirel Pers Commun 2018; 98(1): 665–684.

12.

Hua

Liu

et al . Energy-efficient adaptive slice-based secure data aggregation scheme in WSN. Proced Comput Sci 2018; 129: 188–193.

13.

Ahmad

Hang

Kim

. Design and implementation of cloud-centric configuration repository for DIY IoT applications. Sensors 2018; 18: 474.

14.

Gaur

Scotney

Parr

et al . Smart city architecture and its applications based on IoT. Proced Comput Sci 2015; 52: 1089–1094.

15.

Guan

et al . Achieving efficient and secure data acquisition for cloud-supported internet of things in smart grid. IEEE Internet Things J 2017; 4: 1934–1944.

16.

Zhu

Tan

Zhang

et al . A round-optimal lattice-based blind signature scheme for cloud services. Future Gener Comput Syst 2017; 73: 106–114.

17.

Gao

Cheng

et al . Privacy-preserving Naive Bayes classifiers secure against the substitution-then-comparison attack. Inf Sci 2018; 444: 72–88.

18.

Lin

Huang

et al . A short linearly homomorphic proxy signature scheme. New York: IEEE Access, 2018.

19.

Lin

Yan

Huang

et al . An ID-based linearly homomorphic signature scheme and its application in blockchain. New York: IEEE Access, 2018.

20.

Wei

Zhang

et al . Dynamic fully homomorphic encryption-based Merkle tree for lightweight streaming authenticated data structures. J Netw Comput Appl 2018; 107: 113–124.

21.

Zhang

Xue

et al . An extra-parity energy saving data layout for video surveillance. Multimed Tools Appl 2018; 77: 4563–4583.

22.

Liu

Huang

et al . DivORAM: towards a practical oblivious RAM with variable block size. Inf Sci 2018; 447: 1–11.

23.

Liu

et al . Differentially private Naive Bayes learning over multiple data sources. Inf Sci 2018; 444: 89–104.

24.

Tan

Zhang

et al . A high-performance hierarchical snapshot scheme for hybrid storage systems. Chin J Electron 2018; 27: 76–85.

25.

Wang

Meng

et al . A fog-based privacy-preserving approach for distributed signature-based intrusion detection. J Parall Distrib Comput 2018; 122: 26–35.

26.

Yang

Zhu

Liang

et al . A blockchain-based location privacy-preserving crowdsensing system. Future Gener Comput Syst 2019; 94: 408–418.

27.

Vodel

Hardt

. Data aggregation and data fusion techniques in WSN/SANET topologies—a critical discussion, 2012, https://ieeexplore.ieee.org/document/6412182

28.

Guo

Hong

Zhang

et al . Reliable adaptive data aggregation route strategy for a trade-off between energy and lifetime in WSNs. Sensors 2014; 14: 16972–16993.

29.

Blackert

Gregg

Castner

et al . Analyzing interaction between distributed denial of service attacks and mitigation technologies. In: Proceedings of the DARPA information survivability conference and exposition, Washington, DC, 22–24 April 2003, vol. 1, pp.26–36. New York: IEEE.

30.

Wang

Schulzrinne

. An IP traceback mechanism for reflective DoS attacks. In: Proceedings of the Canadian conference on electrical and computer engineering, Niagara Falls, ON, Canada, 2–5 May 2004, vol. 2, pp.901–904. New York: IEEE.

31.

Pfleeger

. Security in computing. 3rd ed. Upper Saddle River, NJ: Prentice Hall, 2003.

32.

Ozdemir

Peng

Xiao

. PRDA: polynomial regression-based privacy-preserving data aggregation for wireless sensor networks. Hoboken, NJ: John Wiley & Sons Ltd., 2015.

33.

Akila

Sheela

. Preserving data and key privacy in data aggregation for wireless sensor networks. In: Proceedings of the international conference on computing and communications technologies, Chennai, India, 23–24 February 2017, pp.282–287. New York: IEEE.

34.

Ozdemir

Xiao

. Integrity protecting hierarchical concealed data aggregation for wireless sensor networks. Comput Netw 2011; 55(8): 1735–1746.

35.

Zhang

Huang

Wang

et al . Link-based privacy-preserving data aggregation scheme in wireless sensor networks. In: Proceedings of the international conference on industrial IoT technologies and applications, Wuhu, China, 25–26 March 2017, pp.119–129. Cham: Springer.

36.

Elhoseny

Yuan

El-Minir

et al . An energy efficient encryption method for secure dynamic WSN. Secur Commun Netw 2016; 9(13): 2024–2031.

37.

Fang

Wen

et al . CSDA: a novel cluster-based secure data aggregation scheme for WSNs. Clust Comput 2017; (4): 1–12, https://doi.org/10.1007/s10586-017-1195-7

38.

Castelluccia

Mykletun

Tsudik

. Efficient aggregation of encrypted data in wireless sensor networks. 2005; 5(3): 109–117, https://ieeexplore.ieee.org/document/1540992/authors#authors

39.

Eschenauer

Gligor

. A key-management scheme for distributed sensor networks. In: Proceedings of the 9th ACM conference on computer and communications security, Washington, DC, 18–22 November 2002, pp.41–47. New York: ACM.

40.

Karp

Kung

. GPSR: greedy perimeter stateless routing for wireless networks. In: Proceedings of the 6th annual international conference on mobile computing and networking, New York, 2000, pp.243–254, http://www.eecs.harvard.edu/~htk/publication/2000-mobi-karp-kung.pdf

41.

Szewczyk

Ferencz

. Energy implications of network sensor designs. Berkeley Wireless Research Center Report, Berkeley, CA, 2000.

42.

Coman

A, S

ander

Nascimento

. Adaptive processing of historical spatial range queries in peer-to-peer sensor networks. Distrib Parall Database 2007; 22(2–3): 133–163.

43.

Rappaport

. Wireless communications: principles and practice. Upper Saddle River, NJ: Prentice Hall Press, 1996.

Energy-efficient and privacy-preserving spatial range aggregation query processing in wireless sensor networks

Abstract

Keywords

Introduction

Existing research works and their shortcomings

Tree-based algorithms

Cluster-based algorithms

Preliminaries

Attack model

SMART protocol

Energy-efficient and privacy-preserving spatial range aggregation query processing in WSNs—E2PDA

The idea of E2PDA

The design of E2PDA

Grid-based data collection strategy

Grid size design

Method of handling voids

Calculating the next node

Energy consumption analysis

Theorem 1

Proof

Privacy analysis

Analysis of experimental results

Summary

Footnotes

Declaration of conflicting interests

Funding

ORCID iD

References

Energy-efficient and privacy-preserving spatial range aggregation query processing in WSNs—E²PDA

The idea of E²PDA

The design of E²PDA