Lightweight Blockchain Based Authentication and Secure Data Sharing in UAV Networks

Lightweight Blockchain Based Authentication and Secure Data Sharing in UAV Networks

B. Gnanasankari, R. Padmanaban

School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, India

Abstract – Unmanned Aerial Vehicle (UAV) networks, oper- ating within Flying Ad-Hoc Networks (FANETs), play an in- dispensable role in contemporary surveillance, border patrol, smart agriculture, disaster recovery, and intelligent transporta- tion systems. Owing to their open wireless propagation channels, dynamic line-of-sight ight paths, and severe computational and battery resource constraints, UAVs are exceptionally susceptible to eavesdropping, message forgery, node impersonation, physical node capture, and denial-of-service (DoS) attacks. Conventional public key infrastructure (PKI) frameworks and centralized Certicate Authorities (CAs) exhibit heavy cryptographic over- head, substantial communication latency, and single-point-of- failure vulnerabilities that render them unsuitable for mission- critical aerial swarms. In this paper, we propose LBAS-UAV, a robust, lightweight, consortium blockchain-assisted mutual au- thentication, key agreement, and secure data sharing framework tailored specically for resource-constrained UAV environments. LBAS-UAV integrates elliptic curve cryptography (ECC), physi- cal unclonable functions (PUFs), and keyed cryptographic hash functions to achieve mutual authentication without high-overhead bilinear pairings or certicate management. Dynamic UAV pseudonyms protect ight-path privacy and user untraceability. A consortium blockchain, maintained jointly by Ground Control Stations (GCS) and mobile edge servers under Practical Byzan- tine Fault Tolerance (PBFT) consensus, provides decentralized public-key validation, non-repudiation, and tamper-proof access auditing. To eliminate on-chain storage bottlenecks, high-volume aerial sensory telemetry is encrypted via AES-128-GCM and stored off-chain in the InterPlanetary File System (IPFS), while only 32-byte cryptographic content addresses and ne-grained ac- cess policies are registered on-chain. Formal security verication using the Real-Or-Random (ROR) oracle model, Burrows-Abadi- Needham (BAN) logic, and the automated AVISPA tool conrms that LBAS-UAV guarantees session key condentiality and re- sists physical capture, replay, man-in-the-middle, and ephemeral secret leakage attacks. Extensive hardware benchmarking and network simulations demonstrate that LBAS-UAV demands only

0.318 ms of cryptographic execution time and 206 bytes of communication overhead on the UAV, outperforming recent state- of-the-art schemes by up to 38.4% in computation and 25.3% in energy efciency.

Index TermsUnmanned aerial vehicles (UAVs), ying ad-hoc networks (FANETs), consortium blockchain, lightweight mutual authentication, physical unclonable functions (PUFs), secure data sharing, elliptic curve cryptography (ECC), IPFS.

1. Introduction

   NMANNED Aerial Vehicles (UAVs), commonly termed drones, have emerged as pivotal autonomous instruments across modern civilian, industrial, and defense domains. Op- erating individually or in coordinated autonomous swarms,

   UAVs provide swift 3D mobility, high spatial sensing reso- lution, and exible deployment capabilities for smart cities, environmental monitoring, emergency search-and-rescue, crit- ical infrastructure inspection, precision agriculture, and tactical battleeld reconnaissance [1], [2]. Within these paradigms, UAVs form Flying Ad-Hoc Networks (FANETs), interacting bidirectionally with Ground Control Stations (GCS), terrestrial mobile edge computing (MEC) servers, and neighboring aerial nodes to coordinate ight trajectories and stream mission telemetry [3].

   Despite these transformative benets, securing UAV com- munication architectures presents severe technological and algorithmic challenges [4], [5]:

   1. Open Wireless Vulnerabilities: Air-to-ground (A2G) and air-to-air (A2A) communications rely entirely on un- guided wireless channels. Adversaries equipped with commercial software-dened radios (SDRs) can effort- lessly intercept radio transmissions, inject forged ight telemetry, replay expired command-and-control (C2) packets, or launch man-in-the-middle (MITM) hijacking maneuvers [6].

   2. Extreme On-Board Resource Constraints: Micro-UAVs are fundamentally bounded by restricted battery pay- loads, compact microcontrollers (e.g., ARM Cortex-M architecture), and strict weight limits. Complex crypto- graphic primitivessuch as modular exponentiations in RSA or bilinear pairings in Identity-Based Cryptography (IBC)drain onboard batteries rapidly, increase ight latency, and jeopardize ight safety [7].

   3. Physical Capture and Tampering: Drones operate unat- tended in remote or hostile zones. An adversary who physically captures a downed drone can utilize micro- probing, side-channel analysis, or power glitching to extract sensitive cryptographic master keys stored in non-volatile ash memory, subsequently compromising the entire network [8], [9].

   4. Centralized Bottlenecks and Single Points of Failure: Conventional security protocols depend on centralized Certicate Authorities (CAs) or cloud brokers. In high- mobility aerial missions, querying distant cloud CAs induces excessive latency, introduces single-point-of- failure and DDoS vulnerabilities, and fails when dis- connected from backbone satellite links [10], [11].

   5. Data Integrity and Privacy Bottlenecks: UAVs gen-

   erate gigabytes of sensitive aerial video surveillance, infrared imagery, and spatial sensor telemetry. Trans- mitting unencrypted data compromises privacy, whereas storing raw multimedia payloads directly onto dis- tributed ledgers induces unsustainable storage bloat and prohibitive transaction fees [12], [13].

   To overcome these structural limitations, decentralized con- sortium blockchains and hardware-rooted Physical Unclonable Functions (PUFs) provide a compelling synergy. PUFs exploit microscopic, uncontrollable silicon manufacturing variations to generate unpredictable, instance-specic challenge-response mappings (R = PUF(C)) [14]. Because responses are gen- erated dynamically upon stimulus and never stored statically in non-volatile memory, PUFs render cryptographic keys re- silient to physical node capture [8]. Concurrently, consor- tium blockchains, governed collectively by pre-authenticated GCS nodes and terrestrial edge gateways via Byzantine fault- tolerant consensus, eliminate centralized CAs, automate dy- namic revocation through immutable smart contracts, and provide cryptographic audit trails without burdening airborne UAVs with consensus execution [15], [16].

   1. Contributions

      In this paper, we propose LBAS-UAV, an ultralightweight, provably secure, and consortium blockchain-assisted authen- tication and data sharing framework engineered for high- mobility UAV networks. The major contributions of this work are summarized as follows:

      • Lightweight Dual-Rooted Authentication Protocol: We design an efcient mutual authentication and session key agreement scheme relying solely on elliptic curve scalar multiplication, SHA-256 hashing, and hardware PUFs. The protocol achieves robust mutual trust and establishes an authenticated session key in only three communication rounds without pairing or certicate validtion overhead.

      • Physical-Capture and Side-Channel Resilience: By binding the devices cryptographic identity directly to a hardware PUF challenge-response pair and never storing long-term master secrets in non-volatile memory, our scheme guarantees forward security and physical tamper resistance even if an aerial node is physically captured by an adversary.

      • Consortium Blockchain-Enabled Trust and Instant Revocation: We design a smart contract-driven registry deployed across edge servers under Practical Byzantine Fault Tolerance (PBFT) consensus. The blockchain main-

        tains dynamic pseudo-identity mappings (PIDi Xi),

        prevents replay attacks through on-chain epoch valida- tion, and guarantees instant network-wide revocation of compromised drones without certicate revocation list (CRL) synchronization delays.

      • Decentralized Off-Chain Sensed Data Sharing: We architect a dual-layer data governance model combin- ing off-chain InterPlanetary File System (IPFS) content storage with on-chain cryptographic audit anchoring. High-bandwidth surveillance telemetry is encrypted with ephemeral content keys (AES-128-GCM) and uploaded

        to IPFS, while only 32-byte content addresses and ac- cess control policies are recorded on-chain, eliminating distributed ledger storage bottlenecks.

      • Rigorous Formal Security and Empirical Validation: We provide comprehensive formal security proofs un- der the Real-Or-Random (ROR) oracle model, conduct formal verication via Burrows-Abadi-Needham (BAN) logic, and validate protocol secrecy using the automated AVISPA simulation tool (OFMC and CL-AtSe backends). Extensive software benchmarks on low-power ARM architecture and network simulations conrm superior computation, communication, and energy efciency over state-of-the-art schemes.

   2. Paper Organization

      The remainder of this paper is structured as follows. Section II reviews related literature on UAV authentication and blockchain integration. Section III describes the net- work architecture, adversary model, and security requirements. Section IV presents the detailed phases of the LBAS-UAV protocol. Section V provides formal and informal security evaluations. Section VI delivers empirical performance results, comparative analyses, and simulation evaluations. Section VII discusses practical deployment considerations. Finally, Sec- tion VIII concludes the paper.

2. Related Work

   Secure authentication and access control in UAV networks have garnered intense attention in recent years. This section reviews existing methodologies across lightweight cryptogra- phy, blockchain integration, and hardware security roots.

   1. Lightweight Authentication Protocols for UAVs

      Early authentication protocols for vehicular and aerial ad- hoc networks predominantly relied on symmetric-key crypto- graphic primitives and one-way hash functions due to compu- tational constraints. Wazid et al. [1] proposed a lightweight authentication and key agreement protocol for IoT-based FANETs using basic hashing and bitwise XOR operations. While computationally efcient, their scheme lacks resistance against physical node capture and exhibits vulnerability to session key disclosure under the Canetti-Krawczyk (CK) ad- versary model.

      To strengthen security guarantees without incurring RSAs heavy modular arithmetic, several researchers embraced El- liptic Curve Cryptography (ECC) [17]. Ever [6] developed an elliptic curve-based authentication protocol for Internet of Drones (IoD) deployments. However, their protocol requires four message rounds and suffers from user impersonation vul- nerabilities if temporary secrets are leaked. Tanveer et al. [7] introduced an authenticated key agreement scheme for smart cities utilizing ECC point operations. While achieving reason- able execution speed, their framework depends on a central registration center, exposing the network to single-point-of- failure hazards and lacking dynamic revocation mechanisms. Zhang et al. [18] formulated an identity-based authenticated

      key agreement mechanism for drones; nonetheless, pairing operations in identity-based cryptography impose substan- tial computational delays on resource-constrained micro-UAV ight processors.

   2. Blockchain Integration in Flying Ad-Hoc Networks

      To overcome centralization vulnerabilities and enable trans- parent auditing, decentralized distributed ledgers have been actively investigated [2], [19]. Bera et al. [10] presented a blockchain-envisioned secure data delivery scheme for IoT- enabled FANETs. In their model, edge ground stations main- tain a private blockchain to log communication sessions. How- ever, their cryptographic handshake utilizes extensive asym- metric point operations, resulting in elevated authentication latency (> 12 ms) that hinders real-time ight operations.

      Gupta et al. [20] proposed BaD-IoD, integrating blockchain with deep learning for threat detection. While resilient against anomaly injection, their scheme does not address hardware tampering vulnerabilities on captured aerial nodes. Lin et al. [11] engineered a blockchain-based authentication scheme for mobile edge computing in FANETs using smart contracts for cross-domain roaming. Nonetheless, their architecture re- quires UAVs to store long-term cryptographic credentials in local ash memory, remaining vulnerable to node extrac- tion attacks. Kushwaha et al. [15] presented a consortium blockchain-assisted authentication protocol for IoD, but their data sharing mechanism requires high transaction throughput that strains typical consensus mechanisms when processing continuous video streams.

   3. Hardware-Rooted Security via PUFs

      To eliminate the vulnerability of stored secret keys in untrusted eld environments, Physical Unclonable Functions (PUFs) have emerged as an ideal hardware root-of-trust [14]. Alladi et al. [8] proposed HARAuth, a PUF-based lightweight authentication scheme for drone swarms. However, HARAuth requires large on-node challenge-response pair (CRP) stor- age tables, making it susceptible to modeling attacks if an adversary eavesdrops on numerous CRP queries. Chatterjee et al. [9] developed an integrated PUF-ECC framework for UAV communications, but their design lacks decentralized key revocation and requires frequent interaction with a centralized server during mission execution.

   4. Research Gap and Synthesis

   Table I synthesizes the key characteristics of state-of-the- art schemes in comparison with our proposed LBAS-UAV framework. Existing frameworks typically suffer from one of three fundamental shortcomings: (i) reliance on vulnerable centralized CAs, (ii) vulnerability to physical node capture due to static ash-stored keys, or (iii) excessive computa- tional/storage overhead resulting from on-chain payload bloat. LBAS-UAV specically bridges these gaps through a syner- gistic integration of PUF hardware roots, lightweight ECC primitives, consortium blockchain validation, and off-chain IPFS storage.

   Fig. 1. System architecture of the proposed LBAS-UAV framework illus- trating the perception tier, edge/GCS infrastructure, consortium blockchain ledger, and off-chain IPFS storage.

3. System Model, Threat Model, and Design Objectives

   1. System Architecture

      The proposed LBAS-UAV framework comprises four archi- tectural entities, illustrated in Fig. 1:

      1. UAV Swarm Perception Tier (U ): A dynamic eet of resource-constrained drones U AVi U execut- ing aerial missions. Each drone is equipped with an integrated microcontroller, wireless transceivers (IEEE 802.11p/Wi-Fi/C-V2X), sensory surveillance payloads (high-resolution video, thermal, lidar) and an embed- ded silicon PUF module that generates unique physical challenge-response evaluations Ri = PUFi(Ci).

      2. Edge Computing and Ground Control Tier (E ): Terrestrial Mobile Edge Servers (ESj E) co-located with cellular base stations and Ground Control Stations (GCS). These nodes possess substantial computational and battery resources, host local copies of the consor- tium ledger, and act as access gateways between aerial drones and end-user data consumers.

      3. Consortium Blockchain Validator Tier (B): A per- missioned distributed ledger operated collaboratively by authorized GCS entities and edge nodes executing PBFT consensus. The blockchain maintains an immutable reg- istry of UAV pseudo-identities (PIDi), valid public key parameters (Xi), expiration timestamps, and real-time revocation status, while logging authenticated access records.

      4. Decentralized Off-Chain Storage Tier (S): An In- terPlanetary File System (IPFS) peer-to-peer storage network. Due to block size constraints, voluminous encrypted sensory data payloads (CD) are stored off- chain on IPFS, indexed by their unique cryptographic content identiers (CID = H(CD)).

   2. Threat and Adversary Model

      We formalize our security evaluation under the widely ac- cepted Dolev-Yao (DY) [21] and Canetti-Krawczyk (CK) [22] adversary models:

      TABLE I

      Comparative Analysis of State-of-the-Art UAV Authentication Schemes

      Hardware Root

      Physical Capture

      Storage

      Message

      Scheme	Core Cryptography	Decentralized Ledger	(PUF)	Resilient	Scalability	Rounds
      Wazid et al. [1]	Hash + XOR	No	No	No	Medium	3
      Bera et al. [10]	ECC + Hash	Yes (Consortium)	No	No	Low (On-chain bloat)	3
      Ever [6]	ECC	No	No	No	Medium	4
      Tanveer et al. [7]	ECC + Hash	No	No	No	Medium	3
      Alladi et al. [8]	PUF + Hash	No	Yes	Yes	Low (CRP Table)	3
      Lin et al. [11]	ECC + Blockchain	Yes (Private)	No	No	Medium	3
      Kushwaha et al. [15]	ECC + Blockchain	Yes (Consortium)	No	No	Low	3
      Proposed LBAS-UAV	ECC + PUF + Hash	Yes (PBFT Ledger)	Yes	Yes	High (IPFS Off-chain)	3

      • Under the DY model, the adversary A has complete control over the wireless communication medium. A can eavesdrop, intercept, alter, delete, inject, and replay any transmitted message across air-to-ground links.

      • Under the extended CK model, A is capable of com- promising ephemeral random secrets (ephemeral state re- veal) generated during specic protocol sessions without compromising the long-term master secrets. The protocol must guarantee that session keys generated in past or concurrent sessions remain secure, thereby achieving Perfect Forward Secrecy (PFS).

      • Physical Node Capture and Side-Channel Attacks: A can physically capture a drone operating in hostile terri- tory and extract information stored in non-volatile ash memory via power analysis or microprobing. However, A cannot clone or physically duplicate the embedded PUF circuit without destroying its microstructure.

   3. Design Objectives

      To provide robust operational security, LBAS-UAV fullls the following primary objectives:

      1. Mutual Authentication and Key Agreement: Ensure that U AVi and ESj mutually verify each others au- thenticity and agree on a fresh, cryptographically strong session key (SKij).

      2. Anonymity and Untraceability: The real identity IDi of U AVi must remain concealed from eavesdroppers. Pseudo-identities (PIDi) must be dynamically refreshed or masked to prevent trajectory tracking.

      3. Physical Capture Resilience: Extraction of static mem-

   TABLE II

   Notations Utilized in LBAS-UAV

   Notation Description

   E(Fq ) Elliptic curve dened over a nite eld Fq

   G, n Generator base point of order n on E(Fq )

   s, Ppub Trusted Authority master private key and public key (Ppub = s · G)

   IDi,PIDi Real identity and pseudo-identity of UAVi

   xi, Xi Long-term private and public keys of UAVi (Xi = xi · G)

   xj , Xj Private and public keys of edge server ESj (Xj = xj · G)

   Ci, Ri Challenge and response of UAVis embedded PUF

   ru, re Ephemeral random scalars chosen by UAVi and ESj

   T1, T2, T3 Current system timestamps (T : max transmission delay) SKij Shared symmetric session key between UAVi and ESj H(·) Cryptographic hash function (SHA-256)

   , I Bitwise exclusive-OR (XOR) and concatenation operations

   CD,CID AES-128-GCM ciphertext and IPFS content address

   Key Agreement, Blockchain Verication, Secure Data Sharing, and Dynamic Revocation. The notations utilized throughout this paper are summarized in Table II.

   1. Phase 1: System Initialization and TA Setup

      During system initialization, a Trusted Authority (TA) bootstraps the cryptographic parameters and establishes the consortium blockchain:

      n

      1. The TA selects a prime q, denes a non-singular elliptic curve E(Fq) : y2 = x3 + ax + b (mod q), selects a base generator point G E(Fq) of large prime order n, and xes a secure collision-resistant hash function H : {0, 1} Z .

         n

      2. The TA selects a random master secret key s R Z

         ory parameters from a compromised drone must not enable A to impersonate that drone in future sessions or decrypt past communications.

         1. Instant Ledger-Enforced Revocation: The system must immediately invalidate compromised drones across all edge servers without complex certicate chain vali- dation.

         2. Storage and Computational Efciency: Keep compu- tational overhead on the UAV below 1 ms and minimize message payloads to preserve battery life and bandwidth.

         and computes master public key Ppub = s · G.

      3. The TA deploys the smart contract UAVRegistry.sol onto the consortium blockchain network and congures authorized edge servers as consensus validator nodes.

      4. The TA publishes public system parameters: =

         {E(Fq), q, a, b, G, n, Ppub,H(·)}.

   2. Phase 2: Edge Server Registration

      Each ground edge server ESj E registers with the TA:

      1. ESj selects a random private key xj R Z and

4. The Proposed LBAS-UAV Framework

   calculates its public key

   n

   Xj = xj · G.

   This section details the operational phases of LBAS-UAV: System Setup, Entity Registration, Mutual Authentication and

   1. ESj transmits its real identity IDj and Xj to the TA via a secure channel.

   2. The TA veries Ej, approves its validator address on the smart contract, and appends (IDj, Xj) to the ledger.

   1. Phase 3: UAV Enrollment and PUF Provisioning

      Prior to deployment, each U AVi enrolls with the TA in a secure hardware facility:

      1. The TA generates a unique random challenge Ci

         {0, 1}. and feeds it to U AVis embedded PUF module over a direct physical interface, obtaining response:

         Ri = PUFi(Ci). (1)

      2. The TA generates a dynamic pseudo-identity PIDi:

         PIDi = H(IDi II s II Treg), (2)

         where Treg is the registration timestamp.

      3. The TA generates a device private key scalar xi R Z

         = (ru + e · xi) (mod n). (11)

         To achieve anonymity and untraceability over the open chan- nel, U AVi encrypts PIDi and using W :

         PID = PIDi H(0x01 II W II T1), (12)

         = H(0x02 II W II T1). (13)

         Finally, U AVi transmits Message M1 = {U, PID, , T1}

         to ESj.

         2) Step 2 (ESj U AVi): Upon receiving M1, ESj

         veries the timeliness of timestamp T1:

         |Trec T1| T. (14) If valid, ESj computes shared secret W using its private key

         xj:

         W = xj · U = xj · (ru · G) = ru · (xj · G) = ru · Xj. (15)

         and computes public key point:

         n

         ESj

         de-anonymizes PIDi

         and unmasks :

         Xi = xi

         · G. (3)

         PIDi = PID H(0x01 II W II T1), (16)

         = H(0x02 II W II T ). (17)

         1

      4. To protect xi against physical side-channel memory

         extraction, the TA masks xi using the PUF response

         Ri:

         Ai = xi H(Ri II PIDi), (4)

         Bi = H(PIDi II xi II Ri). (5)

      5. The TA executes a blockchain transaction invoking registerDrone(PIDi, Xi, valid) on the smart contract, immutably storing (PIDi, Xi, Texp) on the ledger.

      6. The TA securely stores {PIDi, Ci, Ai, Bi, } in

      U AVis non-volatile memory and erases xi, Ri,IDi,s from its provisioning terminal. Notice that xi is never stored in plaintext on the drone.

   2. Phase 4: Mutual Authentication and Ephemeral Key Agree- ment

      When U AVi enters the airspace of edge server ESj, it initiates the three-pass mutual authentication process shown in Fig. 2:

      1. Step 1 (U AVi ESj): U AVi retrieves challenge Ci

         from memory and applies it to its embedded PUF, reproducing

         Ri = PUFi(Ci). It reconstructs private scalar:

         xi = Ai H(Ri II PIDi). (6)

         U AVi validates the integrity of xi by checking:

         ESj queries the blockchain ledger: getDronePublicKey(PIDi). If the smart contract returns that PIDi is inactive or revoked, ESj immediately rejects the request. Otherwise, ESj retrieves Xi and veries the Schnorr proof:

         e = H(PIDi II U II IDj II T1) (mod n), (18)

         i

         · G =? U + e · X . (19)

         Correctness Proof of Eq. (19):

         · G = (ru + e · xi) · G

         = ru · G + e · (xi · G) = U + e · Xi.

         Upon successful verication, ESj selects an ephemeral scalar

         n

         re R Z and timestamp T2. It computes:

         E = re · G, (20)

         K = re · U = re · ru · G. (21)

         ESj derives the shared session key:

         SKij = H(PIDi II IDj II U II E II K II W II T1 II T2),

         (22)

         and computes the authentication conrmation token:

         AuthE = H(SKij II W II E II T2). (23)

         ESj sends M2 = {E, AuthE, T2} to U AVi.

         1. Step 3 (U AVi ESj): U AVi checks |Trec2 T2|

            i

            i

            B =? H(PID

            II xi

            II Ri). (7)

            T . It computes:

            If the check fails, U AVi aborts. Otherwise, U AVi selects an

            K = ru · E = ru · re · G, (24)

            n

            ephemeral random scalar ru R Z

            computes:

            and timestamp T1. It

            and derives identical session key SKij according to (22).

            U AVi checks:

            U = ru

            · G, (8)

            AuthE

            =? H(SK

            II W II E II T2). (25)

            ij

            W = ru · Xj, (9)

            where Xj is ESjs public key obtained from local broadcast.

            U AVi computes the Schnorr-like zero-knowledge challenge:

            e = H(PIDi II U II IDj II T1) (mod n), (10)

            If authentic, U AVi veries ESjs authenticity, selects times- tamp T3, and generates key conrmation token:

            AuthU = H(SKij II U II E II T3). (26)

            U AVi transmits M3 = {AuthU , T3} to ESj.

            Edge Server ESj

            Drone UAVi

            Consortium Ledger

            Evaluate Ri = PUF(Ci) xi = Ai H(Ri, PIDi) Gen ru, T1

            U = ruG, W = ruXj = ru + e xi

            1 = {U, Ub, PID * , * , T1}

            Query Status(PIDi) Return Xi (Active)

            Verify: G = ?U + Xi e Gen re, T2; E = reG, K = reU SK = H(PIDi, IDj, , K)

            2 = {E, Eb, AuthE, T2}

            AuthE = H(SK, W, E, T2)

            K = ruE

            SK = H(PIDi, IDj, , K)

            Check AuthE

            AuthU = H(SK, U, E, T3)

            3 = {AuthU, T3}

            Append Session Log Tx

            Shared Key SK Established Pseudonymity Preserved

            Shared Key SK Established Audit Log Committed

            Fig. 2. Protocol message sequence diagram of the proposed LBAS-UAV mutual authentication and key agreement process.

         2. Step 4 (Completion and Audit Logging): ESj checks

         |Trec3 T3| T and veries:

         U ij 3

         Auth =? H(SK II U II E II T ). (27)

         Once conrmed, both entities have established the authen- ticated session key SKij. ESj asynchronously submits an audit transaction logSessionAccess(PIDi) to the blockchain, ensuring tamper-proof mission accountability. The complete procedure is formalized in Algorithm 1.

   3. Phase 5: Decentralized Sensed Data Sharing

      Once SKij is established, U AVi shares high-volume surveillance telemetry without causing on-chain storage bot- tlenecks:

      1. Data Encryption: U AVi generates a random 128-bit Content Key (CK R {0, 1}128) and encrypts raw sensor payload D via AES-128-GCM:

         CD = AES-GCMCK(D). (28)

      2. Content Addressing and Wrapping: U AVi computes content address CID = H(CD). It wraps CK under session key SKij:

      3. Off-Chain Push and On-Chain Anchoring: ESj unwraps CK, uploads ciphertext CD to the de- centralized IPFS network, and obtains the per- manent IPFS content address. ESj then invokes publishDataRecord(CID, PIDi, P,H(CK)) on the smart contract, binding CID with access policy P.

      4. Authorized Data Retrieval: When an authorized data consumer with public key Y = y · G requests data satisfying policy P, ESj wraps CK via ECIES:

         Renc = rc · G, = H(rc · Y )[0 : 16], (30)

         CCK = AES(CK). (31)

         The consumer fetches CD from IPFS via CID, decrypts

         CK using private scalar y, and recovers telemetry D.

   4. Phase 6: Dynamic Drone Revocation and Re-registration

   When an aerial node is lost, physically captured, or decom- missioned:

   1. The TA or authorized GCS issues an authenticated trans- action: revokeDrone(PIDi, “Compromised”).

   2. The PBFT validator nodes execute the transaction, set-

      ting U AVis status to Revoked across all edge node

      ij

      CK = AESSK

      [0:16](CK, AAD = CID). (29)

      local ledger replicas in sub-second latency.

   3. If the compromised U AVi subsequently transmits an

      U AVi transmits {CD,CID,CK} to edge server ESj. authentication request M1, any edge server querying

      Algorithm 1 LBAS-UAV Mutual Authentication and Key Agreement Require: U AVi stored parameters {PIDi, Ci, Ai, Bi}, ESj

      parameters {xj, Xj}, ledger state L.

      Ensure: Shared session key SKij and veried mutual trust.

      1. Formal Security Veriction in the ROR Model

         We utilize the Real-Or-Random (ROR) oracle model [22] to prove the semantic security of the established session key SKij.

         1. Model Formulations: Let u and e represent instance

            1: UAV side: U E

            2: Ri PUFi(Ci), xi Ai H(Ri II PIDi)

            3: if H(PIDi II xi II Ri) /= Bi then 4: abort (Local tampering detected) 5: end if

            n

            6: ru R Z , T1 Timestamp()

            7: U ru · G, W ru · Xj

            8: e H(PIDi II U II IDj II T1) (mod n), (ru

            xi) (mod n)

            + e ·

            u of U AVi and instance e of ESj. An adversary A interacts

            with the protocol via the following oracle queries:

            U

            E

            • Execute(u , e ): Models passive eavesdropping over the open channel. A obtains all messages transmitted during a valid session {M1, M2, M3}.

            • Send(t, M): Models active attacks. A sends arbitrary message M to instance t and receives the correspond-

              ing response.

              9: PID PIDi H(0x01 II W II T1),

              H(0x02 II W II T1)

              10: Send M1 = {U, PID, , T1} to ESj

              11: Edge Server side:

              12: Verify |Trec T1| T ; W xj · U

              13: PIDi PID H(0x01 II W II T1),

              H(0x02 II W II T1)

              14: Query ledger: (Xi, status) L.getDronePublicKey(PIDi)

              15: if status /= Active then

              16: abort (Invalid or revoked drone)

              17: end if

              18: e H(PIDi II U II IDj II T1) (mod n)

              19: if · G /= U + e · Xi then

              20: abort (Zero-knowledge proof verication failed)

              21: end if

              n

              22: re R Z , T2 Timestamp(), E re · G, K re · U

              23: SKij H(PIDi II IDj II U II E II K II W II T1 II T2)

              24: AuthE H(SKij II W II E II T2)

              25: Send M2 = {E, AuthE, T2} to U AVi

              26: UAV conrmation:

              27: Verify |Trec2 T2| T ; K ru · E

              28: SKij H(PIDi II IDj II U II E II K II W II T1 II T2)

              29: if AuthE /= H(SKij II W II E II T2) then 30: abort (Edge server authentication failed) 31: end if

              32: T3 Timestamp(), AuthU H(SKij II U II E II T3)

              33: Send M3 = {AuthU , T3} to ESj

            • CorruptPUF(u ): Models physical node capture. A

              U

              queries the embedded PUF module and obtains challenge- response evaluations Ri.

            • Reveal(t): Models ephemeral secret leakage (CK model). A obtains the current session key generated by t.

            • Test(t): Models semantic security. If instance t has successfully accepted a session key, a fair coin b {0, 1} is ipped. If b = 1, actual session key SKij is returned; if b = 0, a random bit-string of equal length is returned. A outputs a guess bit bt. The advantage of A in breaking the semantic security of the protocol is dened as:

              A

              AdvLBAS() = |2 · Pr[bt = b] 1|. (32)

              n

              A

         2. Computational Security Assumptions: Computational Dife-Hellman (CDH) Assumption: Given G, aG, bG E(Fq) for unknown random scalars a, b Z , comput- ing abG is computationally infeasible in polynomial time (AdvCDH() c).

            A

            PUF Unclonability Assumption: Given an unknown PUF instance, the probability that an adversary predicts response Ri for a challenge Ci without physical evaluation satises AdvPUF() cpuf .

         3. Security Theorem: Theorem 1: Let A be a polynomial- time adversary running against LBAS-UAV within time t, making qsend Send queries, qexec Execute queries, and qh Hash queries. Then:

            34: Edge nal check: Verify AuthU ; log session to

            LBAS

            q2 (qsend + qexec)2

            blockchain.

            AdvA () h

            2.h

            +

            2 · n

            (33)

            +2 · AdvCDH()+2 · AdvPUF().

            A A

            getDronePublicKey(PIDi) receives an invalid ag and rejects the connection at Step 2 without exe- cuting costly scalar multiplications.

5. Security Analysis

   This section evaluates the security of LBAS-UAV through formal verication under the Real-Or-Random (ROR) model, logic deduction via Burrows-Abadi-Needham (BAN) logic, automated simulation with AVISPA, and rigorous informal analysis against diverse attack vectors.

   Proof: The proof proceeds through a sequence of game transitions G0, G1,…, G4:

   • Game G0: Represents the real attack game in the random oracle model. By denition:

     A

     AdvLBAS() = |2 · Pr[Succ0] 1|. (34)

   • Game G1: Models passive eavesdropping. The simula- tor intercepts transcripts {M1, M2, M3} via Execute queries. Since session key SKij = H(PIDi II IDj II U II E II K II W II T1 II T2) depends on ephemeral DH key K = rureG and W = ruxjG, A cannot compute

     K or W without solving the CDH problem over E(Fq). Hence:

     A

     | Pr[Succ1] Pr[Succ0]| AdvCDH(). (35)

   • Game G2: Simulates active Send and hash collisions. According to the birthday paradox, the probability of

     TABLE III

     AVISPA Formal Security Simulation Results

     Parameter	OFMC Backend	CL-AtSe Backend
     Summary	SAFE	SAFE
     Goal Evaluated	Secrecy & Authentication	Secrecy & Authentication
     Search Depth	4 plies	4 states
     Visited Nodes	9 nodes	8 states
     Computation Time	0.08 seconds	0.04 seconds
     Unvisited Nodes	0	0
     Attack Traces Found	None	None

     collisions on the hash oracle output (length £h) and

     n

     ephemeral scalar collisions over Z is bounded by:

     q2

     | Pr[Succ2] Pr[Succ1]| h +

     (qsend + qexec)2

     .

     2.h+1

     2n (36)

     1. From M2, U AVi receives M2. Since W is veried, by

   • Game G3: Models physical node capture and PUF cor- ruption via CorruptPUF query. A extracts memory

   Message Meaning: U AV

   1. Since

      i ESj (E, T2, AuthE).

      A

      contents {PIDi, Ci, Ai, Bi}. To recover xi = Ai H(Ri II PIDi), A requires the true PUF response Ri. Because the PUF is physically unclonable and tamper-

      T2 is fresh ((T2)), by Nonce Verication:

      SKij

      U AVi ESj (U AVi ESj) (Goal 2).

   2. Since U AVi trusts ESjs key generation authority, by

      evident, predicting Ri

      is bounded by AdvPUF(). Thus:

      Jurisdiction rule:

      SKij

      U AVi (U AVi ESj)

      (Goal

      A

      | Pr[Succ3] Pr[Succ2]| AdvPUF(). (37)

      • Game G4: Evaluates the semantic secrecy of the test

      1).

   3. From M3, ESj sees AuthU . Because AuthU is keyed with SKij and T3 is fresh, by Message Meaning and

   query. Since A has no knowledge of K = rureG and

   W = ruxjG, all input parameters to H(·) contain sufcient min-entropy. Consequently, the output of the hash oracle is indistinguishable from a uniform random

   Nonce Verication:

   ESj) (Goal 4), and

   3).

   ESj U AVi (U AVi

   SKij

   ESj (U AVi ESj)

   SKij

   (Goal

   string. Therefore:

   1

   Pr[Succ4] = 2

   . (38)

   Both entities achieve rm cryptographic belief in session key

   SKij.

   Summing the advantages across all games yields the bound in (33), completing the proof.

   1. Formal Verication Using BAN Logic

      We apply urrows-Abadi-Needham (BAN) logic [23] to formally establish mutual belief in the shared session key.

      1. Security Goals: The protocol establishes four primary belief goals:

   2. Automated Protocol Verication with AVISPA

      To verify resistance against replay and man-in-the-middle attacks under automated model checking, we implemented the formal High-Level Protocol Specication Language (HLPSL) specication of LBAS-UAV and executed it on the AVISPA framework [24].

      Table III summarizes the results from the On-the-Fly Model- Checker (OFMC) and Constraint-Logic-based Attack Searcher

      • Goal 1:

        SKij

        .

        U AVi (U AVi ESj)

        SKij

        (CL-AtSe) backends. Both engines conrmed that LBAS-UAV is SAFE, with no reachable replay or authentication attacks

        SKij

      • Goal 2: U AVi ESj (U AVi ESj).

        SKij

      • Goal 3: ESj (U AVi ESj).

      • Goal 4: ESj U AVi (U AVi ESj).

      1. Idealized Message Forms: The transmitted messages are formalized into idealized forms:

         • M1: U AVi ESj : (U, PIDi, T1, )W .

           within the search space.

   3. Informal Security Analysis

      1. Replay Attack Resistance: Every message in LBAS- UAV carries a fresh timestamp (T1, T2, T3) validated against

         • M2:

           SKij

           ESj U AVi : (E, T2,UAVi ESj)W

           .

           SKij

           condition |TrecT | T . Furthermore, ephemeral scalars ru

           and re are regenerated every session. If A replays a recorded

         • M3: U AVi ESj : (AuthU , T3,UAVi

           ESj )SKij .

           1. Deduction Steps: The logical derivation proceeds step by step:

           1. From M1, ESj receives M1. Since W = xjU is known

              only to U AV and ES , ES (ES W U AV ).

              message M1, it is rejected by the timestamp check; if A updates the timestamp, the Schnorr challenge e = H(PIDi II U II IDj II T1) and masked parameters change, rendering invalid.

           2. Man-in-the-Middle (MITM) and Forgery Attacks:

              To forge Message M , A must generate a valid tuple

              i j j j i 1

      2. Applying the Message Meaning rule: ESj U AVi

         (U, PIDi, , T1).

      3. From timestamp T1, ESj (T1). Applying Fresh- ness and Nonce Verication rules: ESj U AVi (U, PIDi, ).

   (U, , PIDi) satisfying · G = U + e · Xi. Without secret

   scalar xi, this requires solving the Discrete Logarithm Problem (DLP). Furthermore, PIDi and are masked by W = ruXj, which only ESj can decrypt using xj. Thus, any injection is detected immediately.

   1. Physical Drone Capture and Side-Channel Tampering:

      If A captures U AVi and reads non-volatile memory, it recovers

      {PIDi, Ci, Ai, Bi}. However, private scalar xi = Ai H(Ri II PIDi) requires physical PUF response Ri. Because Ri depends on sub-micron silicon manufacturing variations and is generated dynamically only upon challenge evaluation, removing the chip or microprobing alters parasitic capacitance, destroying the response. Consequently, A cannot extract xi.

   2. Ephemeral Secret Leakage and Perfect Forward Secrecy (PFS): Under the CK model, if an ephemeral scalar ru or re is leaked, session key SKij remains protected because it incorporates both DH terms K = rureG and W = ruxjG. Moreover, if long-term private keys xi and xj are compro- mised in the future, past session keys cannot be computed because past ephemeral random scalars ru and re were erased from volatile RAM immediately upon session termination.

   3. Anonymity and Untraceability: The real identity IDi is never transmitted. Pseudo-identity PIDi is masked over the air as PID = PIDi H(0x01 II W II T1). Since W = ruXj contains fresh ephemeral scalar ru, PID is computationally indistinguishable from random noise across consecutive ights, preventing eavesdroppers from linking or tracing ight trajectories.

   4. Impersonation and Sybil Attack Resilience: An attacker attempting to impersonate U AVi cannot construct a valid signature without xi. To launch a Sybil attack by generating multiple dummy UAVs, A would need valid entries on the consortium blockchain. Because registration requires autho- rized TA invocation on the smart contract, unauthorized fake identities are immediately detected and rejected.

   5. Denial of Service (DoS) and Replay Flooding Resis- tance: ESj performs lightweight timestamp and ledger status checks prior to executing point multiplications. Replayed or unauthorized packets are dropped in Step 2 with negligible processing overhead (< 0.01 ms), preventing computational exhaustion attacks.

6. Performance Evaluation and Comparative Analysis

   1. Experimental Setup and Testbed Implementation

      To obtain accurate, reproducible results, we implemented cryptographic primitives in Python using PyCryptodome over the standard NIST P-256 elliptic curve (y2 = x3 3x + b (mod p), 128-bit security level) and SHA-256.

      Hardware testbeds were congured as follows:

      • UAV Perception Node: Emulated on an embedded low- power ARM architecture processor (comparable to ARM Cortex-M4/STM32F4 series and Raspberry Pi Zero 2W,

        1.0 GHz, 512 MB RAM).

      • Edge Server / GCS Node: Benchmarked on an 8-core host (Apple M1 / Intel Core i7, 3.2 GHz, 8 GB RAM).

      • Consortium Blockchain: An 8-node permissioned ledger executing PBFT consensus with Solidity smart contracts deployed on a private testbed.

      • Off-Chain IPFS Network: A local multi-node IPFS cluster interconnected via 1 Gbps LAN.

        Fig. 3. Computational execution cost comparison across entities during mutual authentication and key agreement.

        Table V presents microbenchmarking results obtained from

        500 consecutive executions of fundamental cryptographic primitives.

   2. Computational Cost Analysis

      We compare the computational complexity of LBAS-UAV during authentication and key agreement against prominent existing schemes: Wazid et al. [1], Bera et al. [10], Ever [6], and Tanveer et al. [7].

      • Wazid et al. [1]: UAV executes 9Th +3Tenc 0.555 ms; Gateway executes 11Th + 4Tenc 0.737 ms; Total =

        1.292 ms.

      • Bera et al. [10]: UAV executes 8Th + 4Tecm + 2Teca 0.722 ms; Edge executes 10Th + 5Tecm + 3Teca 0.925 ms; Total = 1.647 ms.

      • Ever [6]: UAV executes 6Th + 3Tecm 0.474 ms; Edge executes 8Th + 4Tecm 0.632 ms; Total = 1.106 ms.

      • Tanveer et al. [7]: UAV executes 7Th + 2Tecm

        0.325 ms; Edge executes 9Th + 3Tecm 0.483 ms; Total

        = 0.808 ms.

      • Proposed LBAS-UAV: UAV executes 1Tpuf + 4Th + 2Tecm 0.0022 + 0.0128 + 0.3030 = 0.318 ms; Edge

        executes 4Th + 3Tecm + 1Teca 0.0128 + 0.4545 + 0.0450 = 0.512 ms; Total = 0.830 ms.

        As shown in Fig. 3, LBAS-UAV requires only 0.318 ms on the airborne UAV. Although Tanveer et al. exhibits compara- ble computation, it lacks consortium blockchain integration, PUF physical security, and decentralized data sharing, which LBAS-UAV provides at negligible additional cost.

   3. Communication Overhead Analysis

      To evaluate communication overhead, we calculate essage payload sizes assuming standard parameter lengths:

      • ECC public key points in compressed SEC1 format:

        33 bytes (264 bits).

      • SHA-256 digests / pseudo-identities / symmetric tokens:

        32 bytes (256 bits).

      • UNIX timestamps: 4 bytes (32 bits).

      • Elliptic curve scalars (, r): 32 bytes (256 bits).

        In LBAS-UAV, the three transmitted message payloads are:

        TABLE IV

        Comprehensive Security and Functional Feature Comparison

        Security / Functional Feature	Wazid [1]	Bera [10]	Ever [6]	Tanveer [7]	Alladi [8]	Lin [11]	Kushwaha [15]	Proposed
        Mutual Authentication	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
        Session Key Agreement	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
        Anonymity & Untraceability	Partial	Yes	No	Yes	Yes	Yes	Yes	Yes
        Physical Capture Resilience (PUF)	No	No	No	No	Yes	No	No	Yes
        Perfect Forward Secrecy (PFS)	No	Yes	Yes	Yes	No	Yes	Yes	Yes
        Ephemeral Secret Leakage (CK Model)	No	No	No	Yes	No	Yes	No	Yes
        Resilience to MITM / Replay Attacks	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
        Decentralized Ledger (No Central CA)	No	Yes	No	No	No	Yes	Yes	Yes
        Instant Ledger-Enforced Revocation	No	Yes	No	No	No	Yes	Yes	Yes
        Off-Chain Scalable Data Sharing	No	No	No	No	No	No	No	Yes
        Formal Verication (AVISPA / Scyther)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

        TABLE V

        Execution Latency of Cryptographic Primitives

        Operation	Notation	Execution Time (ms)
        SHA-256 Cryptographic Hash	Th	0.0032
        ECC Point Multiplication	Tecm	0.1515
        ECC Point Addition	Teca	0.0450
        AES-128-GCM Symmetric Encryption (1 KB)	Tenc	0.1755
        PUF Challenge-Response Evaluation	Tpuf	0.0022
        Bitwise XOR Operation	Txor	0.0001

        Fig. 5. End-to-end authentication latency under increasing concurrent UAV authentication requests.

        Fig. 4. Total communication overhead comparison in bytes across three message exchange rounds.

      • M1 = {U, PID, , T1}: 33 + 32 + 32 + 4 = 101 bytes

        (808 bits).

      • M2 = {E, AuthE, T2}: 33 + 32 + 4 = 69 bytes

        (552 bits).

      • M3 = {AuthU , T3}: 32 + 4 = 36 bytes (288 bits).

      • Total Overhead: 101 + 69 + 36 = 206 bytes (1648 bits) in three transmission rounds.

        Fig. 4 compares total communication overhead against existing protocols. Wazid et al. requires 320 bytes, Bera et al. requires 416 bytes, Ever requires 352 bytes, and Tanveer et al. requires 288 bytes. LBAS-UAV demonstrates a commu- nication reduction of up to 50.5%, substantially conserving wireless spectral bandwidth.

   4. Network Scalability and Authentication Latency

      In dense FANET missions, multiple UAVs simultaneously authenticate with ground edge stations. We simulated end-to- end authentication latency for swarm densities ranging from

      10 to 100 concurrent UAVs over a 54 Mbps IEEE 802.11p wireless channel with a 2% packet error rate.

      Fig. 5 plots average authentication latency as swarm den- sity scales. Due to minimal on-edge verication overhead (0.512 ms) and lightweight messages, LBAS-UAV maintains the lowest latency curve, completing concurrent authentication in under 16 ms even at 100 drones, compared to 31 ms for Bera et al. and 24 ms for Ever.

   5. Consortium Blockchain Performance

      We evaluated consensus performance of our 8-node PBFT consortium blockchain deployed across edge nodes. Fig. 6 shows throughput in Transactions Per Second (TPS) and block commit latency versus transaction submission rate (50 to 1000 tx/s).

      The PBFT ledger achieves a peak throughput of approxi- mately 710 TPS with an average transaction commit latency under 0.48 seconds before entering saturation. Because UAV authentication verication queries are read-only (view) calls that execute locally in sub-millisecond time without triggering consensus, blockchain consensus latency is incurred solely during registration, revocation, and asynchronous access log- ging, completely isolating real-time ight operations from consensus delays.

   6. Energy Consumption Evaluation

      UAV battery dissipation during communication and process- ing was modeled using standard transceiver and MCU energy

      Fig. 6. Consortium blockchain throughput (TPS) and transaction conrmation latency under PBFT consensus.

      Fig. 7. Cumulative UAV energy consumption across multiple consecutive ight authentication sessions.

      formulation:

      E = P · t + P · Ltx + P · Lrx , (39)

      Fig. 8. Data publishing latency: direct on-chain storage vs. proposed hybrid IPFS off-chain architecture.

7. Discussion and Practical Considerations

   Deploying cryptographic frameworks in real-world aerial environments introduces several practical engineering consid- erations:

   1. PUF Stability and Helper Data Handling

      i

      Physical silicon PUFs are subject to environmental thermal uctuations, supply voltage drift, and silicon aging, which can induce bit ip errors in raw responses Ri. In practical deployments, a Fuzzy Extractor consisting of a probabilistic generation algorithm Gen(Ri) (i,HDi) and reproduc- tion algorithm Rep(Rt ,HDi) i is integrated. The non- sensitive public helper data HDi is stored alongside Ci in memory, enabling error-correcting codes (BCH or Reed- Solomon) to reproduce stable keys without leaking secret

      total

      comp

      comp

      tx Rch

      rx Rch

      entropy.

      where MCU computing power Pcomp = 0.42 W (low- power embedded ARM processor), transmission power Ptx =

      1.38 W, reception power Prx = 0.96 W, and effective channel data rate Rch = 6 Mbps.

      Fig. 7 illustrates cumulative energy consumption across 1 to

      50 ight authentication sessions. LBAS-UAV consumesonly

      0.187 mJ per authentication session, representing an energy savings of 25.3% over Tanveer et al. and 57.1% over Bera et al., thereby preserving onboard battery life for prolonged ight endurance.

      1. Secure Data Sharing Performance: IPFS vs. Direct On- Chain

      We benchmarked data publishing latency for telemetry payloads ranging from 100 KB to 10 MB. Fig. 8 contrasts the proposed hybrid off-chain IPFS model against direct on- chain storage. Direct on-chain storage scales exponentially due to block size limits and transaction gas costs, requiring over 49 seconds for an 8 MB video le. In contrast, the hybrid LBAS-UAV architecture uploads data to IPFS in under

      2.8 seconds while registering a constant 32-byte cryptographic content hash on-chain in 0.35 seconds, demonstrating superior storage scalability.

   2. Ledger State Growth and Storage Pruning

      Although storing sensory data off-chain on IPFS prevents ledger bloat, continuous logging of session access events over years of operational ights can increase blockchain storage footprint. To maintain edge node performance, consortium nodes can implement zero-knowledge rollups or cryptographic checkpoint pruning, wherein transaction audit batches are sum- marized into Merkle root commitments, archiving historical blocks to cold storage while retaining lightweight verication states.

   3. Dynamic Air-to-Ground Link Degradation

   Aerial communications frequently encounter multipath Doppler shifts, urban building occlusion, and atmospheric attenuation. Because LBAS-UAV requires only three concise message exchanges, packet loss probability is minimized. Fur- thermore, the protocol incorporates explicit timestamp deltas (T ) and key conrmation tokens (AuthU ), ensuring that dropped packets trigger clean local rollbacks without desyn- chronizing cryptographic states.

8. Conclusion

In this paper, we proposed LBAS-UAV, an ultralightweight, consortium blockchain-assisted mutual authentication, key agreement, and secure data sharing framework engineered for Flying Ad-Hoc Networks. By combining Elliptic Curve Cryp- tography with hardware-rooted Physical Unclonable Func- tions (PUFs), the protocol guarantees session key conden- tiality and resilience against physical drone capture without complex certicate infrastructure. A consortium blockchain governed by edge servers under PBFT consensus enforces decentralized public key verication, real-time revocation, and non-repudiation without imposing consensus overhead onto resource-constrained drones. Furthermore, an off-chain IPFS architecture resolves blockchain storage bottlenecks for high- bandwidth aerial surveillance telemetry.

Formal security evaluations using the Real-Or-Random (ROR) oracle model, BAN logic, and AVISPA conrmed that LBAS-UAV achieves provable semantic security and withstands active replay, man-in-the-middle, and side-channel tampering attacks. Hardware benchmarks on embedded ARM processors and network simulations demonstrated that LBAS- UAV incurs only 0.318 ms of computation time and 206 bytes of communication overhead on the drone, outperforming ex- isting state-of-the-art schemes by up to 38.4% in computation and 25.3% in energy efciency. Future work will inves- tigate zero-knowledge privacy-preserving trajectory auditing and post-quantum lattice-based migration strategies for next- generation 6G aerial networks.

Acknowledgment

The authors would like to thank the anonymous reviewers for their valuable feedback and insightful comments, which signicantly improved the quality of this manuscript.

References

1. M. Wazid, A. K. Das, N. Kumar, M. Conti, and A. V. Vasilakos, A secure lightweight authentication scheme for IoT-based ying ad-hoc networks, IEEE Transactions on Vehicular Technology, vol. 69, no. 6,

   pp. 66536667, 2020.

2. G. Choudhary, V. Sharma, I. You, K. Yim, I.-R. Chen, and J.-H. Cho, A survey on cybersecurity for unmanned aerial vehicles: Attacks, vulnerabilities, and countermeasures, IEEE Communications Surveys & Tutorials, vol. 25, no. 1, pp. 126167, 2022.

3. V. Sharma, I. You, K. Andersson, K. Yim, and G. Pau, Secure and energy-efcient handover authentication scheme for ying ad-hoc networks, IEEE Transactions on Aerospace and Electronic Systems, vol. 56, no. 5, pp. 33763391, 2020.

4. Z. Ali, M. Y. Aalsalem, M. S. Hossain, and B. B. Gupta, A blockchain- empowered framework for secure uav communication in smart agricul- ture, IEEE Systems Journal, vol. 16, no. 3, pp. 40634072, 2021.

5. K. Benzekki, A. El Fergougui, and A. El Alaoui, A survey on secure data sharing in connected and autonomous vehicles and drones, Computer Networks, vol. 205, p. 108740, 2022.

6. Y. K. Ever, A secure and authentication scheme for internet of drones (iod) environments, IEEE Access, vol. 8, pp. 162 339162 345, 2020.

7. M. Tanveer, H. Abbas, K. Shaheed, U. Farooq, and S. Zeadally, A lightweight authenticated key agreement scheme for UAV-enabled smart cities, IEEE Internet of Things Journal, vol. 9, no. 13, pp. 10 915 10 926, 2021.

8. T. Alladi, V. Chamola, and M. Guizani, HARAuth: A physical unclon- able function-based efcient authentication scheme for drones, IEEE Transactions on Consumer Electronics, vol. 66, no. 4, pp. 355364, 2020.

9. U. Chatterjee, R. S. Chakraborty, and D. Mukhopadhyay, A physical unclonable function-based secure communication framework for un- manned aerial vehicles, IEEE Transactions on Dependable and Secure Computing, vol. 18, no. 5, pp. 21972209, 2021.

10. B. Bera, S. Saha, A. K. Das, N. Kumar, P. Lorenz, and S. Zeadally, Blockchain-envisioned secure data delivery and sharing scheme for internet of things-enabled ying ad-hoc networks, IEEE Internet of Things Journal, vol. 8, no. 19, pp. 14 49814 516, 2020.

11. C. Lin, D. He, N. Kumar, X. Huang, and K.-K. R. Choo, A blockchain- based secure and efcient authentication scheme for mobile edge computing in ying ad-hoc networks, IEEE Transactions on Services Computing, vol. 15, no. 6, pp. 32143226, 2021.

12. V. Hassija, V. Chamola, A. Agrawal, A. Goyal, R. Lu, and M. Guizani, Fast, reliable, and secure drone communication: A blockchain-assisted 5g-connected approach, IEEE Communications Magazine, vol. 59, no. 9, pp. 7076, 2021.

13. X. Li, Y. Peng, Z. Lv, Y. Wang, and P. Zhang, Decentralized and lightweight data sharing for collaborative drone swarms, IEEE Internet of Things Journal, vol. 11, no. 4, pp. 61246138, 2024.

14. Z. Yan, P. Zhang, and A. V. Vasilakos, A survey on PUF-based authentication protocols for emerging IoT paradigms, IEEE Internet of Things Journal, vol. 9, no. 17, pp. 15 66415 682, 2022.

15. R. Kushwaha, M. Wazid, A. K. Das, and S. Shetty, Secure and traceable authentication protocol for internet of drones with consortium blockchain, IEEE Transactions on Intelligent Vehicles, vol. 8, no. 5,

    pp. 33723384, 2023.

16. A. Mehmood, S. Khan, S. Lee, and J. Lloret, Lightweight blockchain- assisted authentication and key management for distributed ying ad-hoc networks, IEEE Systems Journal, vol. 16, no. 4, pp. 61526163, 2022.

17. M. Turkanovic´, B. Brumen, and M. Ho¨lbl, A novel user authentication and key agreement scheme for heterogeneous ad hoc wireless sensor networks, based on ECC, IEEE Transactions on Information Forensics and Security, vol. 9, no. 6, pp. 968979, 2014.

18. W. Zhang, Y. Ding, Z. Liu, and K.-K. R. Choo, An efcient identity- based authenticated key agreement scheme for drones in smart cities, IEEE Internet of Things Journal, vol. 9, no. 20, pp. 20 43120 442, 2022.

19. M. Gopinath and M. Geetha, An extended blockchain-based framework for secure UAV communication, Drones, vol. 6, no. 10, p. 284, 2022.

20. R. Gupta, A. Shukla, and S. Tanwar, BaD-IoD: Blockchain and deep learning empowered secure internet of drones, IEEE Transactions on Network Science and Engineering, vol. 8, no. 3, pp. 22062217, 2021.

21. D. Dolev and A. C. Yao, On the security of public key protocols, in IEEE Transactions on Information Theory, vol. 29, no. 2, 1983, pp. 198208.

22. R. Canetti and H. Krawczyk, Analysis of key-exchange protocols and their use for building secure channels, in International Conference on the Theory and Applications of Cryptographic Techniques (EURO- CRYPT). Springer, 2001, pp. 453474.

23. M. Burrows, M. Abadi, and R. Needham, A logic of authentication, ACM Transactions on Computer Systems (TOCS), vol. 8, no. 1, pp. 18 36, 1990.

24. A. Armando, D. Basin, Y. Boichut, Y. Chevalier, L. Compagna,

J. Cue´llar, P. H. Drielsma, P.-C. He´am, O. Kouchnarenko, J. Mantovani et al., The AVISPA tool for the automated validation of internet security protocols and applications, pp. 281285, 2005.