Decentralized Spectrum Sharing: A Proof-of-Concept Evaluation

10th Workshop on Communication Networks and Power Systems (WCNPS 2025) Decentralized Spectrum Sharing: A Proof-of-Concept Evaluation João Gabriel ∗, Edilson Filho †, Wilker Feitosa †, Carlos F. M. e Silva † Thiago Nóbrega ∗, Francisco R. P. Cavalcanti † Federal University of Campina Grande (UFCG), Brazil∗ Federal University of Ceará (UFC), Brazil† Abstract—Wireless communication is an important tool for economic growth and fostering digital inclusion in society. As the need for wireless services continues to rise, optimizing the use of underutilized frequency bands has become mandatory. This makes effective spectrum sharing a condition to meet novel communication needs. Nevertheless, the existing spectrum management approaches, often centralized, face burdens regarding reliability, and establishing trust among users. This work introduces the Decentralized TV White Spaces Management System (DWS-MS), whose central hypothesis is that a decentralized, blockchain-based system can effectively manage wireless spectrum across extensive geographical regions. In addition, this work presents a real-scale test conducted using a region in Brazil to validate its practical applicability. Keywords—spectrum database, TV White Spaces, decentralized systems, blockchain I. Introduction The electromagnetic spectrum has become an important resource, acting as an enabler for telecommunications and economic growth [1]. Effective spectrum sharing is necessary to address the growing demand for wireless communication; it enables the reuse of underutilized frequency bands, expands the capacity of wireless networks, and fostering social and economic development. Moreover, it plays a relevant role in promoting digital inclusion, particularly in remote or underserved regions like Brazil [2], by facilitating access to broadband connectivity with lower costs and without the need for specialized personnel. One example of dynamic spectrum sharing is the TV White Spaces (TVWS) concept, which taps into the unused portions of the broadcast television spectrum and repurposes them for broadband connectivity without harming incumbent TV services. The TVWS refers to the unused portions of the broadcast television spectrum that can be utilized for other wireless communication purposes without interfering with primary users [2]. TVWS offers a model for achieving coexistence in shared spectrum bands using databasedriven spectrum management. Traditional methods of spectrum management, characterized by exclusive licensing and uncoordinated unlicensed access, have faced limitations in meeting the high demands of wireless devices, applications, and other clients. In response to these challenges, coordinated spectrum systems have been developed to improve spectrum utilization efficiency by allowing different users, with potentially different priorities, to share the same frequency bands [3]. To enable dynamic spectrum coordination, georeferenced databases have been proposed. These databases allow devices to query and identify which frequency bands are available for use in a specific geographic area, thereby minimizing interference and enhancing coexistence among heterogeneous wireless technologies [3]. In essence, they support dynamic spectrum access by providing real-time, location-based information on spectrum availability. However, the implementation of these systems face several key problems, which include the reliability and security of spectrum allocations, scalability of spectrum management systems given the computational complexity, the necessity of trust among the clients, and difficulties in achieving harmonious coexistence between various wireless technologies operating in shared bands [2], [4]. Centralized spectrum database implementations have limitations that can affect their effectiveness and adoption. For instance centralized databases create single points of failure, where system outages or cyberattacks can disrupt spectrum access for entire regions [5]. Moreover, centralized spectrum database models often raise trust concerns among operators and Internet Service Providers (ISPs), who may fear biased access to spectrum information [4]. In practice, this lack of trust encourages different organizations to develop and maintain their own local spectrum databases, each implementing distinct rules, formats, and update frequencies. This fragmented landscape results in data inconsistencies and difficulties in coordinating spec-979-8-3315-7075-0/25/$31.00 ©2025 IEEE 10th Workshop on Communication Networks and Power Systems (WCNPS 2025) trum sharing, particularly across regional or national boundaries. Consequently, the absence of a unified and transparent coordination mechanism undermines the overall efficiency and fairness of spectrum access and hampers large-scale adoption of shared spectrum systems. These databases generally demand sophisticated calculations for the majority of functions, including spectrum availability determination, interference prediction, coexistence control, and policy application [2], [1]. Therefore, the larger the number of secondary users, the volume of the spectrum allocations to be managed, and the sophistication of sharing algorithms, the greater the computational overhead of the system, consequently influencing both performance and scalability [1]. With these concepts in mind, the Decentralized TV White Spaces Management System (DWS-MS) was proposed as a proof of concept (PoC) for implementing decentralized TV White Spaces databases, addressing the limitations of conventional centralized approaches. The DWS-MS decentralized architecture tackles the shortcomings of traditional centralized spectrum databases by incorporating blockchain technology, bringing both opportunities and challenges for spectrum management. The blockchain based operations provide immutability and auditability, enabling all spectrum transactions and allocations to be transparently recorded on a shared ledger. This transparency fosters user trust and confidence in the fairness of the allocation process, as every participant can independently verify system actions and spectrum assignments [4]. The resulting distributed trust model mitigates the dependence on centralized authorities, facilitating broader participation in spectrum sharing markets, particularly by smaller operators and ISPs who can now engage on verifiable conditions. Furthermore, the use of blockchain with the Proof-of-History (PoH) [6] consensus mechanism guarantees that transactions are collectively validated. Nevertheless, the distributed nature of the system may slow decision making processes and complicate considerations regarding scalability limits. Our evaluation indicates that while the decentralized approach may not necessarily excel over traditional systems in every metric, it serves as a viable alternative in scenarios that prioritize transparency, auditability, and distributed trust, especially as blockchain technologies advance and optimize for performance and critical applications. The study presented in this paper seeks to validate the operational capacity of the DWS-MS architecture in continental-scale settings while guaranteeing trust among the clients, accuracy at spectrum allocation, and capacity to answer the client query at a reasonable time. For this, we test the scalability through load testing and system profiling in a real brazilian environment. The remainder of the paper is organized as follows. In Section 2, we show some essential concepts to understand the DWS-MS architecture and discuss related literature. Section 3 details the DWS-MS architecture and blockchain integration. Section 4 outlines the experimental methodology and main results. Section 5 discusses the DWS-MS deployment requirements to support large-scale environments. Finally, section 6 highlights the main conclusions. II. Preliminaries and Related Work Efficient spectrum management is essential for the deployment of TV White Spaces (TVWS), as improper allocation could lead to harmful interference with primary users such as television broadcasting or air trafic control bands. To prevent this, geolocation spectrum databases ensures that secondary users access available spectrum without disrupting licensed operations [7]. These databases must continuously collect and process vast amounts of regulatory, geospatial, and propagation data, including transmitter locations, signal power, terrain profiles, and real-time environmental conditions. Research on spectrum databases and decentralized management frameworks has gained traction with the increasing demand for efficient, secure, and scalable spectrum access solutions. In the following we show related works, highlighting how each addresses spectrum management efficiency, security, privacy, and flexibility. Mfupe and Mekuria [8] developed a centralized Geo-location White Space Spectrum Database (GLWSDB) for managing TVWS in South Africa, comparing its performance with a commercial system and field measurements. Feng et al. [9] presented the design and implementation of WhiteNet, a databaseassisted multi-access points system operating on TVWS. WhiteNet aimed to dynamically access vacant TV spectrum with the help of a geolocation database. They also proposed a low-overhead distributed spectrum allocation algorithm to allocate spectrum among multiple access points (APs). Chen et al. [10] examine decentralized spectrum sharing mechanisms incorporating database support within dynamic spectrum access environments, proposing a distributed management approach that lacks Blockchain integration. The work of Sumithra and Shirly [11] proposes an auction-based mechanisms in cognitive radio networks. Although their approach 10th Workshop on Communication Networks and Power Systems (WCNPS 2025) enables dynamic allocation, it is mainly specific to cognitive radio environments. DWS-MS differs by allowing flexible allocation mechanisms and ensuring decentralized trust through Blockchain, enhancing applicability across multiple contexts. Unlike previous approaches that focus on centralized spectrum management or specific regulatory environments, the proposed DWS-MS introduces a decentralized and spectrum-agnostic framework. Traditional methods [8], [9], rely on centralized databases, limiting flexibility and increasing dependency on regulatory entities. III. Proposed Solution The DWS-MS architecture, has three interconnected modules, the Application Programming Interface (API),Spectrum Board Module(SBM), and Propagation Module(PM) modules, as shown in Figure 1. The API, serves as the entry point, receiving spectrum queries requests using the RFC 7545 standard [12], which standardizes the communication message exchanges between a user White Space Device (WSD) and a TVWS database. The SBM, acts as the database storing and processing spectrum access requests made by devices and users, leveraging blockchain technology to manage and authorize spectrum use in a secure, transparent, and auditable manner. Therefore SBM eliminates the dependency on a central authority, as it is available in every node of the blockchain network. Finally we have the PM, responsible for performing the physical calculations that define signal coverage and ensure that the new usage does not cause interference with existing services. Blockchain Network API GATEWAY USER (WSD) PROPAGATION MODULE (PM) Blockchain Node Blockchain Node Blockchain Node RFC 7545 JSON-RPC HTTP/ JSON-RPC Figure 1: DWS-MS Architecture The SBM handles spectrum queries using the PoH consensus algorithm [6]. PoH’s high throughput and low latency are critical for meeting the strict responsetime requirements of real-time spectrum management over large areas [13]. The SBM’s functionality is delivered through five interconnected smart contracts: the Location Service (managing geographical data), the Primary Transmitter Service (integrating primary user data), the Device Service (regulating equipment), the Spectrum Service (checking channel availability), and the Authentication, Authorization and Accounting Service (AAA) (handling user authentication and authorization). Locations are represented using pixels, which serve as discrete, square-shaped units of geographic space. These pixels enable spatial mapping and efficient management of data concerning spectrum availability and usage within their defined boundaries. The pixel can be see as the square units in Figure 2. The dimension and area covered by a single pixel can vary depending on the specific requirements of the dynamic spectrum sharing system, the density of primary and secondary users, and the desired granularity for spectrum management in a given region. For instance, a pixel might represent an area as small as 1000m2 (e.g., a 31m x 31m square) in densely populated urban environments or as large as 10,000 m2in more sparsely populated areas. The PM calculates the coverage of primary systems by determining key parameters for each geographic pixel. Using processed georeferenced data from the SBM blockchain, it computes the TV noise floor, desired power, undesired power, and signal-tointerference-and-noise ratio (SINR). To estimate the aforementioned parameters the PM employs the Longley-Rice propagation model [14] to simulate signal propagation across varied terrain types, ensuring detailed coverage analysis. For secondary applications requiring reduced computational effort, the Extended-Hata model [15] is alternatively used. With the SINR calculation applied on a propagation model, it outputs a Radio Environmental Map (REM), showing the SINR for each location and for each channel. After the calculation effected by the PM, the SBM returns to the API the response that states which spectrum is available. A. Propagation details The objective of the propagation module is to identify viable transmission zones for secondary users-such as small ISPs. Figure 2 illustrates this process. The broadcast area of primary users (TV stations) are shown in red. To avoid interference with TV broadcasts, secondary users must determine the maximum transmission power permitted at each location. To identify the channels available for secondary use, do not interfere with the primary user telecomunication services, in the pixels (locations) we use a SINR threshold δ. For example, the secondary user will not cause interference with the antennas A, B, and C in the model 10th Workshop on Communication Networks and Power Systems (WCNPS 2025) n A n Bn C  2 km 4 km Figure 2: Spectrum board logical representation represented by Figure 2, but it cannot increase its signal power, or it will interfere with the antenna C signal. Notice that, to avoid interference, the system calculates the distance from the secondary transmitter to the edge of the primary coverage. This pixel-level analysis is computationally intensive, requiring signal quality estimation across the entire spectrum board. This process, related to defining the available pixels and the distance between the signal areas’ borders, is computationally demanding because it necessitates assessing the primary users’ estimated coverage and signal quality for every pixel in the spectrum board. To understand how this is made, see Algorithm 1. The algorithm to get the available channels for the WSD in a specific location can be divided into two phases. In the first phase, the objective is to identify, at the WSD location, which channels are free for secondary use. It accepts three inputs: the S, containing the spectrum board; p, the specific pixel coordinates at which the WSD is requesting service and δrepresenting the SINR threshold. It then initializes an empty list of candidate channels, represented by Ω, that will eventually hold only those channels deemed free at that location. Next, the procedure enters a nested loop structure that iterates through each location l∈ L, transmitter t∈ T , and channel c∈ C. For each combination, the algorithm invokes getData (using the blockchain) to retrieve the signal power and stores the value in the sinrMap[l][c]array. Once all transmitters and channels have been processed and the SINR map for the pixel is fully populated, the algorithm proceeds to a second pass over the channels (line 8). The acceptance criterion is applied: the SINR must fall below a predefined threshold δ. Algorithm 1: Identification of TVWS candidate channels at the device’s pixel location Input: S: Spectrum board p: WSD pixel location δ: SINR threshold Output: Set of candidate TVWS channels Procedure: getAvailableChannels(S,p,δ) 1Ω← ∅ 2sinrMap ← ∅ 3foreach location l∈ L do 4foreach transmitter t∈ T do 5foreach channel c∈ C do 6channelData ←getData(S, p, t, c) 7sinrMap[l][c]←channelData.power 8foreach channel c∈ C do 9sinrValue ←sinrMap[l][c] 10 if sinrValue < δ then 11 Ω.add({c, sinrValue}) 12 return Ω Whenever the condition is satisfied, the channel and its corresponding SINR value are appended to the Ωlist, which contains tuples (channel, sinr)for every channel that is safe to use at that one pixel. In the next phase, the Algorithm 2 takes the shortlist of candidate channels (Ωlist) to determine, the maximum geographic radius within which the WSD client can safely transmit without causing harmful interference to primary users. The algorithm has five inputs: the S,p,Ωand δalready explained; Rmax which represents the maximum search radius the algorithm should search. The algorithm first initializes an empty set Σto store the results. It then iterates through all pixels lwithin a maximum radius Rmax of the WSD’s position p. For each pixel and for every candidate channel c, it calculates the SINR based on a simulated transmission from pto l. If the SINR is below the threshold δ, the algorithm updates Σ[c]to track the maximum distance at which channel cis still safe for use. By the end of execution, the algorithm will have determined, for each candidate channel, the maximum allowable distance from the WSD at which transmission is still permitted. The total computation complexity of the algorithm is quadratic in relation to the pixel size; the proof can be seen in Appendix A. 10th Workshop on Communication Networks and Power Systems (WCNPS 2025) Algorithm 2: Determination of the maximumdistance border for each candidate channel Input: S: Spectrum board p: WSD pixel location Ω: List of candidate channels δ: SINR threshold Rmax: Maximum search radius (km) Output: Σ: List of tuples {channel,maxDistance } Procedure: calculateChannelBorders(S,p,Ω,δ, Rmax) 1Σ← ∅ 2L←getNearbyP ixels(S, p, Rmax) 3foreach location l∈Ldo 4foreach channel c∈Ωdo 5power ←getData(S, l, c, p) 6SINR ←power −calculateNoiseF loor(c) 7if SINR < δ then 8distance ←calculateDistance(p, l) 9Σ[c]←max(distance, Σ[c]) 10 return channelBorders IV. System Validation To evaluate whether the proposed decentralized architecture is capable of supporting spectrum management at continental scale, we conducted a series of experiments. The validation process was structured around two dimensions: A. Query responsiveness under load — to evaluate the system’s ability to handle multiple simultaneous spectrum availability requests, while respecting the 60 second response time constraint [16], [17]; B. System computation efficiency — to evaluate the computational cost and execution time of the spectrum availability calculations at the pixel level; In our evaluation we selected a test region in Brazil encompassing approximately 3,500 km2, covering the municipalities of Fortaleza, Caucaia, and Quixadá. This region was chosen not only for its mix of urban and rural characteristics, but also for its high density of primary transmitters: 1,011 TV transmitters reported by the Brazilian Telecommunications Authority (ANATEL). For context, this number is nearly four times greater than the total number of transmitters in the entire country of Portugal, which has 262 [18]. This density and geographic diversity make the region a suitable proxy for evaluating the system’s behavior under continental-scale conditions. The experiments were executed in a controlled environment using a server equipped with an Intel(R) Xeon(R) Silver 4216 CPU at 2.10 GHz and 64 GB of DDR4 RAM. A pixel size of 1 km2was adopted for the spectrum board, which offers a challenge regarding computational granularity and geographic resolution for the population density and propagation characteristics of the test region. Each experiments were executed 300 times to ensure statistical significance. A. Query responsiveness under load To evaluate a single server’s ability to handle multiple requests, we conducted load testing, increasing the number of concurrent users from 1 to 50. Figure 3 presents box plots with the response time for different numbers of simultaneous users. It shows that while response times increase with the number of users, they remain under the 60-second threshold until we exceed 20 users. Beyond this point, the growth becomes exponential, causing response times to exceed 60 seconds. Additionally, we can also estimate future costs in terms of response time as we scale the number of users, exponential fit (highlighted in red). 1 3 5 10 15 20 50 Simultaneos Users 10 20 30 40 50 60 Respone Time in sec Simultaneos Users Response Time Exponencial Fit Figure 3: Response time per concurrent users Regarding the Blockchain performance, we observed that the component did not exhibit a significant increase in CPU or memory usage, with the entire impact being absorbed by the API. In contrast, the most demanded blockchain node maintained a low CPU usage (31%) even under heavy load, demonstrating its resilience. In conclusion, the API handle up to 20 simultaneous requests without errors. Beyond this point, response times start to exceed the 60-second limit. 10th Workshop on Communication Networks and Power Systems (WCNPS 2025) B. System computation efficiency To better understand the performance bottlenecks within the system, we conducted a series of tests targeting the primary modules of the DWS-MS: API, SBM, and Propagation. The goal was to profile the system by isolating and measuring the execution time per module. Figure 4 shows the distribution of response times per module. Notably, two actions dominate the total response time: Pixel Available Channels and Validate Channel Border. These actions represent significant processing overhead due to the computational complexity involved in checking spectrum availability and ensuring compliance with channel boundaries, detailed in Algorithms 1 and 2. Message validation Pixel Identification Pixel Availabel Channels Validade Ch Border Max TX Power Estimation 0.0 0.5 1.0 1.5 2.0 2.5 Delta Time (seconds) System Response time by load component TVWS Layer API Blockchain Propagation Figure 4: System response time by component The median response time for Pixel Available Channels, primarily handled by the blockchain layer, was approximately 0.5 seconds. Similarly, the validation of channel borders, managed by the API layer, exhibited a median response time of around 2.0 seconds. These two steps account for a substantial portion of the overall processing time, representing approximately 87% of the total system execution time. V. Conclusion This work addressed the limitations of traditional centralized spectrum management, particularly regarding transparency, cost, and user trust, by proposing the DWS-MS. The DWS-MS fosters trust through the inherent transparency of the blockchain. Moreover, our experiment provides a continental-scale evaluation to confirm the system’s capability to manage numerous concurrent users efficiently over a wide area, validating its practical applicability. Future research directions include enhancing the DWS-MS by supporting a broader spectrum range - e.g., 6G bands. Acknowledgments This work is part of the project Implementing TV Whitespace (TVWS) for internet access in Brazil: challenges and opportunities which was funded by The Government of the United Kingdom of Great Britain and Northern Ireland acting through the Foreign, Commonwealth & Development Office (FCDO) and by the Núcleo de Informação e Coordenação do Ponto BR (NIC.BR). Appendix In Algorithm 1, the Available Channel Identification, the outermost loop iterates over all geographic locations (pixels) L, requiring O(L)operations. For each location, a secondary loop processes all TV transmitters T, resulting in O(L·T)total iterations. Within this structure, a third loop evaluates all frequency channels C, performing constant-time validations, which contribute O(L·T·C)operations. OperationsPhase1 =L×T×C =L×(αL)×C(since T=αL) =αCL2(L2dominates the complexity) =O(L2) The number of transmitters is proportional to the pixel number, so the overall complexity of phase 1, as can be seen above, is O(L2). In the second phase (Maximum Operational Distance Calculation), there exists one loop over Lpixels and one loop over Cchannels per pixel. This results in a total of O(L), since Cis constant. The total computational complexity combining both phases is O(L2+L), this is equal to O(L2), since it is dominated by the quadratic term. References [1] S. Bhattarai, J.-M. J. Park, B. Gao, K. Bian, and W. Lehr, “An overview of dynamic spectrum sharing: Ongoing initiatives, challenges, and a roadmap for future research,” IEEE Trans. Cogn. Commun. Netw., vol. 2, no. 2, pp. 110–128, 2016. [2] E. Filho, “Building trust in spectrum sharing,” in Networking Data Integrity and Manipulation in Cyber-Physical and Communication Systems, R. Herrero, Ed. Cham, Switzerland: Springer, 2025, pp. 1–37. [3] P. Lin, J. Jia, Q. Zhang, and M. Hamdi, “Dynamic spectrum sharing with multiple primary and secondary users,” IEEE Trans. Veh. Technol., vol. 60, no. 4, pp. 1756–1765, 2011. [4] L. Perera, P. Ranaweera, S. Kusaladharma, S. Wang, and M. Liyanage, “A survey on blockchain for dynamic spectrum sharing,” IEEE Open Journal of the Communications Society, vol. 5, pp. 1753–1802, 2024. [5] M. Banafaa, I. Shayea, J. Din, M. H. Azmi, A. Alashbi, Y. I. Daradkeh, and A. Alhammadi, “6g mobile communication technology: Requirements, targets, applications, challenges, advantages, and opportunities,” Alexandria Engineering Journal, vol. 64, pp. 245–274, 2023. 10th Workshop on Communication Networks and Power Systems (WCNPS 2025) [6] B. Lashkari and P. Musilek, “A comprehensive review of blockchain consensus mechanisms,” IEEE Access, vol. 9, pp. 43 620–43 652, 2021. [7] ETSI, “Reconfigurable radio systems (RRS); use cases for operation in white space frequency bands,” European Telecommunications Standards Institute, TR 102 907, Oct. 2011, version 1.1.1. [8] L. Mfupe and F. Mekuria, “Geo-location white space spectrum databases: Models and design of South Africa’s first dynamic spectrum access coexistence manager,” KSII Trans. Internet Inf. Syst., vol. 8, no. 11, pp. 3810–3836, Nov. 2014. [9] X. Feng, J. Zhang, and Q. Zhang, “Database-assisted multi-AP network on tv white spaces: Architecture, spectrum allocation and AP discovery,” in Proc. IEEE Int. Symp. Dyn. Spectrum Access Netw. (DySPAN), 2011, pp. 265–276. [10] J. Chen, C. Yin, Z. Ma, and L. Zhang, “Database-assisted distributed spectrum sharing in dynamic spectrum access systems,” IEEE J. Sel. Areas Commun., vol. 35, no. 2, pp. 417–427, 2017. [11] D. S. S. Sofia and E. A. S. Edward, “Auction based game theory in cognitive radio networks for dynamic spectrum allocation,” Comput. Electr. Eng., vol. 86, 2020. [12] V. Chen, S. Das, L. Zhu, J. Malyar, and P. McCann, “Protocol to access white-space (PAWS) databases,” RFC 7545, Internet Engineering Task Force, May 2015. [13] A. Yakovenko, “Solana: A new architecture for a high performance blockchain,” Solana Labs, Whitepaper, 2018, [Online]. Available: https://solana.com/solana-whitepaper.pdf. [14] M. Hata, “Empirical formula for propagation loss in land mobile radio services,” IEEE Trans. Veh. Technol., vol. 29, no. 3, pp. 317–325, Jul. 1980. [15] G. A. Hufford, A. G. Longley, and W. A. Kissick, “A guide to the use of the its irregular terrain model in the area prediction mode,” Nat. Telecommun. Inf. Admin. (NTIA), Tech. Rep. 82100, Apr. 1982. [16] S. Bhattarai, J.-M. Park, B. Gao, K. Bian, and W. Lehr, “An overview of dynamic spectrum sharing: Ongoing initiatives, challenges, and a roadmap for future research,” IEEE Transactions on Cognitive Communications and Networking, vol. 2, no. 2, pp. 110–128, 2016. [17] A. Attar, S. A. Ghorashi, M. Sooriyabandara, and A. H. Aghvami, “Challenges of real-time secondary usage of spectrum,” Computer Networks, vol. 52, no. 4, pp. 816–830, 2008. [18] ANACOM, “Emissores | TDT,” TDT Telecom Portugal, [Online]. Available: https://tdt.telecom.pt/emissores. [Accessed: Apr. 25, 2025].