PaperlayerView exact revision on arXiv

Source-linked AI summary

A Survey on Digital Twins: Architecture, Enabling Technologies, Security and Privacy, and Future Prospects

Yuntao Wang, Zhou Su, Shaolong Guo, Minghui Dai, Tom H. Luan, Yiliang Liu

arXiv:2301.13350v1cs.CRcs.NI

TL;DR

IoDT promises coordinated information exchange and mission execution across physical entities and virtual twins, but its distributed, intelligent, and semantic architecture creates significant security and privacy challenges. This paper surveys IoDT architecture, enabling technologies, threats, countermeasures, and future directions. It provides a comprehensive synthesis of working principles, security and privacy issues, and potential defenses for the emerging IoDT paradigm.

  • Problem

    IoDT’s distributed structure, information-centric routing, autonomous intelligence, and semantic communications create security and privacy challenges that conventional approaches cannot solely resolve.

  • Method

    The paper conducts an in-depth survey of IoDT architecture, communication modes, enabling technologies, security and privacy threats, countermeasures, and future research directions.

  • Results

    The survey synthesizes IoDT working principles, a distributed cyber-physical architecture, threat taxonomy, security challenges, state-of-the-art countermeasures, and future research directions.

  • Takeaways & Limitations

    The survey aims to provide a thorough understanding of IoDT and inspire further research toward efficient and secure IoDT systems.

Abstract

from arXiv · show

By interacting, synchronizing, and cooperating with its physical counterpart in real time, digital twin is promised to promote an intelligent, predictive, and optimized modern city. Via interconnecting massive physical entities and their virtual twins with inter-twin and intra-twin communications, the Internet of digital twins (IoDT) enables free data exchange, dynamic mission cooperation, and efficient information aggregation for composite insights across vast physical/virtual entities. However, as IoDT incorporates various cutting-edge technologies to spawn the new ecology, severe known/unknown security flaws and privacy invasions of IoDT hinders its wide deployment. Besides, the intrinsic characteristics of IoDT such as \emph{decentralized structure}, \emph{information-centric routing} and \emph{semantic communications} entail critical challenges for security service provisioning in IoDT. To this end, this paper presents an in-depth review of the IoDT with respect to system architecture, enabling technologies, and security/privacy issues. Specifically, we first explore a novel distributed IoDT architecture with cyber-physical interactions and discuss its key characteristics and communication modes. Afterward, we investigate the taxonomy of security and privacy threats in IoDT, discuss the key research challenges, and review the state-of-the-art defense approaches. Finally, we point out the new trends and open research directions related to IoDT.

I. INTRODUCTION

The Internet of digital twins (IoDT) interconnects physical entities and virtual twins for information sharing, synchronization, cooperation, and composite insights, while introducing substantial security and privacy challenges. This survey reviews IoDT architecture, enabling technologies, threats, defenses, and future directions.

  • IoDT Concept and Benefits: IoDT connects massive physical entities with virtual twins that exchange information, synchronize statuses, and cooperatively perform missions through intra-twin and inter-twin communications.Aggregated data from connected twins can produce composite insights across many physical entities.
  • Enabling Technologies: IoDT combines AI, semantic communications, cloud-edge computing, SAGIN, and blockchain to support high-fidelity twins, low-latency transmission, computing, networking, and trust establishment.Blockchain contributes decentralized ledgers, distributed consensus, and trust-free smart contracts for data and value exchange.
  • Security and Privacy Challenges: Security and privacy risks arise across pervasive data collection, massive data sharing, and critical-infrastructure operation because twin data is delay-sensitive and mission-critical.Digital-twin data traverses multiple networks, software systems, and applications during service provision.
  • Security and Privacy Challenges: IoDT’s autonomous intelligence, decentralized structure, information-centric routing, and semantic communications intensify existing vulnerabilities and create new security-service tradeoffs.Applications must balance service latency, system overhead, and security provision under differing QoS requirements.
  • Survey Scope: This survey reviews IoDT architecture, communication modes, enabling technologies, seven categories of security and privacy threats, countermeasures, and open research directions.The seven threat perspectives are data, authentication, communication, privacy, trust, monetization, and cyber-physical security.

C. Organization of Our Survey

The survey is organized around IoDT architecture and its physical-cyber components, followed by enabling technologies, security and privacy issues, countermeasures, and future research. Its architecture links physical entities and virtual twins through an IoDT engine using bidirectional data and feedback flows.

  • Survey Organization: The survey first introduces IoDT architecture, communication modes, key characteristics, and enabling technologies, then covers threats, countermeasures, future directions, and conclusions.Sections III and IV address security and privacy issues and state-of-the-art countermeasures from seven aspects.
  • Architecture of Internet of Digital Twins: IoDT comprises physical entities, digital twins with virtual assets, and an IoDT engine linking physical and cyber spaces through input big data and output feedback.The engine maintains the bidirectional connection between physical entities and their digital twins.
  • Physical Entities: Physical entities include sensing, control, hybrid, and infrastructure types that gather data, execute cyber-layer instructions, or provide grid, networking, computing, caching, and storage capacities.Sensing entities include IoT sensors, smart meters, and wearable devices; control entities execute relevant actions.
  • Digital Twin: Digital twins are software representations of real-world entities, systems, processes, or abstractions that interact and synchronize with physical counterparts in real time.Twins can be deployed on cloud or edge servers and linked through synchronized private connections.
  • Internet of Digital Twins: IoDT interconnects multiple sub-IoDTs so virtual twins can share information, synchronize with physical objects, cooperate on missions, and aggregate distributed data into composite insights.The IoDT engine uses IoT, AI, semantic communication, and blockchain to create, maintain, and update digital representatives.
  • Enabling Technologies: IoT supplies sensing, networking, and computing infrastructure; AI supports high-fidelity real-time simulations; semantic communication transmits task-essential meaning; blockchain supports provenance and trust.Semantic communication targets low-latency, low-overhead exchanges between physical entities and twins and among twins.

B. Communication Modes of Digital Twins

IoDT uses inter-twin communication for cooperation among twins and intra-twin communication to bridge each physical entity with its digital twin. These modes support information sharing, synchronization, and coordination across physical and virtual entities.

  • Inter-Twin Communication: Inter-twin communication enables data access and sharing between digital twins, supporting collaboration for physically distant entities.Twins in cloud or edge environments can exchange information beyond physical communication-range limitations.
  • Intra-Twin Communication: Intra-twin communication builds private, bidirectional data-flow links between a physical entity and its digital twin.The physical entity supplies real-time data, while the twin returns feedback and smart decisions.
  • Intra-Twin Communication: Intra-twin synchronization requirements range from real-time (∼millisecond) through near real-time (∼second) to delay-tolerant (∼minute) services.Different applications therefore require different synchronization levels.
  • Illustrating Example: An IoDT traffic-scheduling example combines inter-twin traffic sharing with intra-twin semantic data flows for prediction and autonomous sensor decisions.Twin UAVs obtain information from twin UAVs and vehicles, while each virtual UAV learns from its physical counterpart.

C. Key Characteristics of Internet of Digital Twins

IoDT is designed as a flexible information-sharing system whose digital twins can act autonomously and retrieve information by content rather than by host. Its decentralized, delay-sensitive setting motivates information-centric communication mechanisms.

  • Autonomous Intelligence: Digital twins can proactively seek valuable information from relevant twin nodes and autonomously synchronize with their physical entities after authorization.These actions can occur without notifying or being instructed by physical entities.
  • Decentralized Structure: IoDT lacks a central server for massive heterogeneous twin nodes, while spontaneous and delay-sensitive exchanges make centralized routing costly.Centralized networking may introduce unnecessary data hops and extra latency.
  • Information-Centric Routing: Information-centric routing retrieves demanded information based on interests rather than the identity of a specific data source.Examples include publish/subscribe and named data networking paradigms.
  • NDN: In NDN, hierarchical names guide interest packets, and routers use content stores, pending interest tables, and forwarding information bases.Routers can return cached content or process outstanding and new content interests.
  • Pub/Sub: In publish/subscribe routing, publishers advertise content through brokers and subscribers request content objects through their local brokers.The paradigm uses flat names containing a topic ID and a unique content ID.

4) Semantic Communications:

Semantic communication transmits task-relevant meanings rather than complete bit sequences, reducing communication burden in IoDT. However, heterogeneous twins and semantic processing introduce interoperability, data-quality, synchronization, routing, and model-security concerns.

  • Semantic Communications: Semantic communication extracts and transmits task-relevant information, filtering irrelevant content to reduce bandwidth consumption.This approach is intended for both intra-twin and inter-twin communications.
  • Interoperability: Heterogeneous physical entities, software, interfaces, communication modes, data types, and twin-production methods create severe interoperability problems.The heterogeneity spans both hardware and software components.
  • Communication Modes: Intra-twin semantic communication supports end-to-end multimodal transmission between physical entities and twins, while inter-twin communication supports multi-agent cooperation among virtual twins.Intra-twin settings may combine image, video, voice, and text from multiple sensors; inter-twin settings use shared and private knowledge bases.
  • Semantic Adversarial Attacks: Semantic adversarial attacks perturb raw data or channel-side processing during inference, exploiting the vulnerability of DNN-based semantic communication.Such attacks can occur in both intra-twin and inter-twin interactions.
  • Semantic Data/Knowledge Poisoning: Semantic data or knowledge poisoning injects contaminated samples into raw data or knowledge bases to manipulate model training and degrade performance.The attack can occur during communication between twins and physical entities or among twins.
  • Data-Related Threats: Semantic IoDT systems face threats including cache pollution, data tampering, low-quality data, desynchronization, model inconsistency, and data or content poisoning.These threats can affect routing, synchronization, model training, and digital-twin reactions.

B. Threats to IoDT Authentication •

Authentication-related IoDT threats include identity abuse, unauthorized access, compromised components, rogue infrastructure, malicious virtual assets, and privilege escalation. These threats can expose sensitive information or alter digital-twin systems.

  • Identity Threats: Impersonation attacks exploit authentication flaws to assume legitimate identities and extract credentials or security parameters.The threat applies to both intra-twin and inter-twin interactions.
  • Unauthorized Access: Unauthorized data access exposes fine-grained personal and sensitive information collected for intelligent digital-twin services.Malicious users or providers may use accessed information for targeted advertising and precision marketing.
  • Knowledge-Base Threats: Unauthorized knowledge-base access can expose or tamper with public and agent-private knowledge bases, affecting semantic communication and privacy.Multi-agent communication distinguishes between publicly accessible and agent-private knowledge bases.
  • Compromised Components: Backdoors inserted into devices or software can interrupt normal operations, cause malfunctions, or leak information.The threat includes compromised components or code introduced by malicious or disreputable manufacturers.
  • Rogue Infrastructure: Rogue devices, servers, and virtual assets can replace legitimate resources, modify digital twins, or provide footholds for further attacks.Rogue servers may control digital replicas of physical entities, while rogue virtual assets can facilitate transitive attacks.
  • Privilege Escalation: Privilege escalation lets insiders or external attackers gain illicit access to protected resources by exploiting software and network weaknesses.The resulting access can leak highly sensitive user data.

C. Communication-Related Threats in IoDT •

IoDT communication channels and cooperation mechanisms face attacks that expose semantic information, consume resources, manipulate exchanges, or interrupt twin operations.

  • Semantic eavesdropping: Semantic eavesdropping can recover semantic information under poor channel conditions and expose aspects of users’ real data distributions.A shared decoder can help eavesdroppers decipher semantic information even when channel quality is low.
  • Flooding attacks: Flooding attacks send large volumes of messages or unresolved interests to disable twin nodes or overload network resources.Interest flooding specifically targets CPU or memory through thousands of unresolved content requests.
  • Communication manipulation: Man-in-the-middle attackers insert themselves between connected entities, potentially altering messages and controlling their conversation.Victims may be made to believe they are communicating directly with each other.
  • Identity attacks: Sybil attacks exploit one node to operate multiple identities and gain influence in decentralized IoDT interactions.A majority of influence can undermine authority in reputable systems, as illustrated by a 51% attack in Bitcoin.
  • Denial of service: Denial-of-service attacks exhaust constrained device resources or exploit network and application layers, interrupting digital-twin simulation and prediction.Possible causes include jamming, blackhole, sinkhole, wormhole, flooding, and malware injection.

D. Privacy Threats to IoDT •

IoDT privacy threats arise from extensive personal-data collection, privileged access, AI-model processing, and difficult accountability across digital-twin services.

  • Personal data: Pervasive personal-data collection at high granularity and synchronization frequency creates opportunities for crimes and misuse of sensitive twin data.The data supports accurate digital clones but expands exposure of personal information.
  • Insider threats: Insiders can exploit privileges to extract credentials, access digital twins, steal stored personal information, and conduct cyber espionage.The threat spans intra-twin and inter-twin interactions.
  • Regulatory compliance: Digital-twin service providers must obtain user grants and protect privacy when collecting, storing, transmitting, or processing personal data under regulations such as GDPR.The requirement applies throughout big-data analysis for service delivery.
  • Model privacy: Model aggregation can leak training data when semi-honest cloud or edge servers use plaintext gradients to reconstruct original samples.Generative adversarial networks are identified as one reconstruction technique.
  • Model privacy: Model delivery and deployment expose risks of model theft, parameter-based privacy inference, tampering, and backdoor insertion.The threat occurs while storing, delivering, and deploying trained global models.
  • Inference attacks: Membership inference attacks infer sensitive training samples from AI-model outputs without direct access to the training data.Training samples are not irreversibly hidden by conversion into model parameters.
  • Inference attacks: Knowledge or model inversion attacks reconstruct private training datasets from representations encoded in AI models.The paper distinguishes white-box attacks with structural access from black-box attacks through open APIs.
  • Misuse and accountability: Authorized providers may disclose or adversaries may sell personal data, while copying and complex service cycles make accountability difficult.The passage links misuse concerns with difficulty tracing misbehaving entities and enforcing accountability quickly.

E. Trust Issues in IoDT •

IoDT trust issues span the reliability of exchanged data, fairness of transactions and participation, transparency of markets, digital-asset provenance, and cyber-physical consequences.

  • Data trustworthiness: Untrusted twins may exchange falsified information, while adversaries can modify or replace real-time data synchronized between physical entities and twins.These concerns affect both intra-twin and inter-twin interactions.
  • Transaction fairness: Inter-twin transactions can involve sellers providing falsified models or services and buyers refusing payment, creating trust and fairness issues.The threat is described as transaction fraud in digital-twin data exchanges.
  • Participation: Free-riding physical entities or twins may consume services without contributing, such as sharing redundant information during collaborative model training.The setting is an open and untrusted IoDT.
  • Market transparency: Opaque resource or knowledge trading can produce disputes over prices and service quality in public IoDT markets.Heterogeneous entities need transparent sharing and trading arrangements for collaborative tasks.
  • Asset provenance: Digital assets are easy to copy across platforms, and multiple ownership forms complicate proving provenance and separating ownership from use rights.The passage identifies both singly and collectively owned assets.
  • Intellectual property: Digital-twin models face intellectual-property threats through explicit resale or implicit extraction methods such as pruning and distillation.Model theft for profit can hinder practical deployment of digital-twin services.
  • Cyber-physical security: Because IoDT bridges cyber and physical spaces, attacks can target both digital systems and physical processes.Physical damage and cyber compromise are presented as two lines of attack.
  • Cyber-physical security: Physical destruction of devices or servers can create a single point of failure and disrupt optimization and monitoring services in cyberspace.The threat directly affects normal digital-twin operations.

IV. SECURITY AND PRIVACY COUNTERMEASURES IN IODT

Countermeasure research addresses IoDT data synchronization, consistency, security, resilience, and service optimization through blockchain, verification, learning, optimization, and decentralized architectures.

  • IoDT Data Security, Resilience & Consistency: Provable data possession and consortium blockchain support trusted time-state verification, integrity checking, anonymity, and unforgeability for multi-source synchronization.The cited design uses tag verification to prevent legitimate virtual spaces from being framed and is analyzed under RSA assumptions.
  • IoDT Data Security, Resilience & Consistency: Sustainable data collection methods jointly address reveal delay, information loss, data fidelity, and energy constraints in dynamic IoDT settings.The approach targets long-term collection while preserving useful information.
  • IoDT Data Security, Resilience & Consistency: A Dolev–Yao-based security architecture and state replication mechanism target synchronization, software, network-isolation, and DoS-resilience issues.A PLC proof-of-concept assesses components of the proposed design.
  • IoDT Data Security, Resilience & Consistency: Blockchain-based twinchain frameworks address sensitive-information leakage, data manipulation, data reliability, quantum resistance, and transaction confirmation.Permissioned blockchains, auction pricing, and DT-DPoS are also used to secure and coordinate digital-twin transactions.
  • IoDT Data Security, Resilience & Consistency: Related studies apply digital twins to synchronization, vehicular traffic, industrial IoT, edge offloading, VR resource allocation, and resilient air-ground networking.Reported objectives include lower latency, improved transaction throughput, resource optimization, and enhanced data resilience.

B. IoDT Authentication & Access Control

IoDT authentication and access control research spans vehicular, smart-grid, and general IoDT settings, combining signatures, blockchain, AI, encryption, and smart contracts. Related work also extends protection toward intrusion detection, situational awareness, and digital-twin placement.

  • IoDT Authentication in IoV: Vehicular IoDT authentication addresses both intra-twin and inter-twin communications using group signatures, secret handshakes, and proxy ring signatures.These approaches target conditional anonymity, unlinkability, cybertwin migration, and mutual authentication between vehicles and roadside edge nodes.
  • Blockchain for IoDT Authentication in IoV: Blockchain-assisted vehicular authentication stores public keys in transparent ledgers, supports privacy-preserving group authentication, and uses GANs for twin-risk forecasting.The reported goal is to mitigate impersonation threats while assisting IoDT authentication.
  • AI and Blockchain for IoDT Authentication in Smart Grid: Smart-grid authorization combines AI-based feature prediction and optimization with blockchain-enforced automatic access control.The authorization platform uses transparent blockchain ledgers to support access-policy enforcement.
  • Access and Usage Control in IoDT: Attribute-based encryption and smart contracts support application-dependent, fine-grained data access and usage control in IoDT.The cited schemes include KP-ABE and CP-ABE, while smart contracts stipulate access and usage policies.
  • Intrusion Detection, Situational Awareness, and Placement: IoDT security research also covers terminal-to-terminal intrusion detection, traffic prediction, situational awareness, and digital-twin placement under dynamic resource constraints.Placement studies formulate edge-association problems, with numerical results reporting improved convergence rates in complex network scenarios.

D. Privacy Countermeasures in IoDT

IoDT privacy countermeasures use blockchain, federated learning, and privacy-computing techniques to protect data across physical systems, digital twins, and service lifecycles. Federated learning keeps local data at clients while supporting decentralized model construction and resource scheduling.

  • Blockchain for Privacy Preservation in IoDT: Blockchain-based approaches address privacy preservation for labeling and tracking physical objects while supporting real-time synchronization between digital twins and edge systems.The cited work applies blockchain to privacy preservation in IoDT and uses DTEN integration to support synchronization.
  • Federated Learning for Privacy Preservation in IoDT: Federated learning trains models locally without uploading private data, balancing user privacy protection with decentralized big-data utilization for IoDT models.Federated analytics keeps personal data from being shared within digital twins, while reported results demonstrate accuracy and convergence against benchmark schemes.
  • Federated Learning for Privacy Preservation in IoDT: Digital-twin federated learning has been applied to distribution estimation and resource scheduling, with reported accuracy, convergence, low latency, and security objectives.The cited resource-scheduling algorithm, SAINT, is reported to outperform state-of-the-art algorithms in simulation.
  • Other Technologies to be Explored: Differential privacy, secure multi-party computing, and homomorphic encryption are identified as additional privacy-computing options across digital-twin service lifecycles.These techniques are presented as lessons for privacy protection beyond blockchain and federated learning.
  • Trust Management in IoDT Services: Trust management supports high-fidelity digital twins by evaluating the trustworthiness of sensory and processing data, while blockchain records collaboration activities for traceability and provenance.Blockchain-based schemes use decentralized storage, smart contracts, and trusted data tracing in digital-twin data management.
  • Trust-Based Model Aggregation in IoDT Services: Trust-based federated aggregation can combine blockchain privacy preservation with GAN-empowered differential privacy noise applied to local model parameters.The cited FedTwin scheme is asynchronous and defines differential privacy during local training.

F. Provenance, Governance & Accountability in IoDT

IoDT governance and accountability research responds to centralized architectures, single points of failure, security-performance trade-offs, stealthy attacks, congestion, and the need to protect physical systems. Proposed mechanisms span permissioned blockchains, deep learning, game theory, incentives, and digital-twin forensics.

  • Blockchain for IoDT Provenance and Governance: Permissioned blockchain is used to build decentralized and flexible digital-twin governance architectures that address poor flexibility and single-point-of-failure risks in cloud-based systems.Blockchain-based designs also target auditability and traceability for critical data.
  • Blockchain for IoDT Provenance and Governance: A two-layer blockchain framework applies proof-of-authority consensus to transparent quality-control verification in a digital-twin hemp supply chain.The associated digital-twin model uses stochastic simulation for risk management with dynamic evolution and spatial-temporal causal interdependencies.
  • Deep Learning for IoDT Governance: Deep learning supports secure IoDT governance in intelligent transportation, with CNN and support vector regression used to improve prediction accuracy.The cited simulation reports a security prediction accuracy of 90.43%.
  • Game-Theoretical IoDT Governance: Game-theoretical approaches address stealthy estimation attacks, attack defense, service congestion governance, and long-term incentive design.One cited threat involves attackers learning defense strategies to alter digital-twin state estimation without detection.
  • Incentive Design for IoDT Governance: Credit-based incentives and contract-theoretical mechanisms are used to optimize edge resource allocation and mitigate latency-critical service congestion under dynamic demands.The contract mechanism incorporates individual rationality, incentive compatibility, and delay sensitivity.
  • Digital Twin for Protecting Physical Systems/Infrastructures: Digital twins support physical-system protection and live or postmortem forensics through replication, historical-data analysis, and simulation modes.These modes help mirror current industrial-control-system states, analyze attack histories, and investigate attack origins and lateral movements.
  • Economic and Social Effects in Defenses: Existing CPS digital-twin services emphasize accuracy and processing speed, while economic and social costs are often ignored.A green AI-enabled surveillance framework is proposed to reduce resource consumption and motivate reusable-device participation.

V. FUTURE RESEARCH DIRECTIONS

Future IoDT research targets scalable and seamless service orchestration, heterogeneous networking, interoperability, and explainable AI. The survey frames these directions alongside its architecture, threat taxonomy, countermeasure review, and open research agenda.

  • Cloud-Edge-End Orchestration: Cloud-edge-end orchestration is proposed to handle terminal growth, service interruptions, and mobility beyond access-point coverage by collaboratively establishing service function chains.The architecture comprises cloud, edge, and end tiers for enhanced QoS.
  • Space-Air-Ground Integrated Networking: SAGIN offers space, air, and ground subnetworks for ubiquitous coverage and ultra-wide-area broadband access, while heterogeneous resources and protocols create security and privacy challenges.Dynamic network environments may affect service performance across these heterogeneous tiers.
  • Interoperability: IoDT interoperability requires information exchange across digital twins and between cyber and physical spaces across hardware, software, protocols, interfaces, and operating systems.Open challenges include developing new standards spanning these dimensions.
  • Explainable AI: Explainable AI is important for IoDT because AI supports high-fidelity twin evolution, adaptable semantic communications, security situation awareness, and regulatory functions.Interpretable machine learning has been applied to digital-twin updating using interpretable physical and mathematical models.
  • Survey Scope and Open Directions: The survey introduces a distributed IoDT architecture, explains inter-twin and intra-twin information flows, and reviews supporting technologies, security/privacy threats, challenges, countermeasures, and future directions.Its stated goal is to provide an in-depth understanding of IoDT working principles and inspire further research.
Loading 2301.13350v1…