Science AI Journal
Computer Science · Research topic

Open research questions in Software-Defined Networks and 5G

56 unresolved questions extracted from the limitations and future-work sections of 204 Software-Defined Networks and 5G papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • Traditional routing protocols lack robust mechanisms to ensure data integrity and defend against cyber threats. The dense device population density, dynamically changing network topologies, and changing traffic patterns in smart city environments have presented numerous technical challenges to traditional routing protocols.

    A blockchain enabled IoT routing framework improving security and performance in smart cities · 2026 · DOI

  • Further validation of the hybrid NDT model on larger networks. Investigation of the applicability of the model to other types of networks.

    Learning a network digital twin as a hybrid system · 2026 · DOI

  • The need for a hybrid NDT model that captures cell-specific properties and handover phenomena. The need for an annealing optimization-based learning algorithm for identifying and continuously improving the hybrid NDT model.

    Learning a network digital twin as a hybrid system · 2026 · DOI

  • Delayed Congestion Reaction. Unnecessary rate throttling. Limited awareness of ECMP routing decisions.

    AN ENHANCED ECN-BASED CONGESTION MITIGATION WITH TRAFFIC REROUTING FOR CLOS DATA CENTER NETWORKS · 2026 · DOI

  • The paper identifies a research gap in the understanding of the potential applications of 5G Standalone Core Networks in IoT, M2M, and smart infrastructure development. The study highlights the need for further research on the governance, security, and interoperability of 5G SA-enabled smart infrastructure.

    ROLE OF 5G SA CORE NETWORKS IN ENABLING IOT, M2M, AND SMART INFRASTRUCTURE DEVELOPMENT ACROSS SAUDI ARABIA’S VISION 2030 PROJECTS · 2026 · DOI

  • The diversity of heterogeneous artificial-intelligence work creates a recurring architectural problem. The need to treat each narrowing decision as an explicit projection. The requirement to remain operational enough to execute work and constrained enough not to become a second policy authority.

    NIMBUS Connect Reference Architecture · 2026 · DOI

  • In this research article, we proposed MOTA-SVB, a novel SDN-based controller placement approach for vehicular networks that integrates vehicle trajectory prediction in terms of physical distance with respect to RSUs, with a multi-objective optimization algorithm for controller placement. The optimisation algorithm minimizes the physical distance between vehicles and RSUs, reduces communication delay, and minimizes the number of controllers placed, while ensuring efficient load balancing and achieving BFT. Extensive simulations demonstrate that our approach significantly outperforms an existing state-of-the-art method by optimizing controller placement, reducing communication delay (by up to 17%), computation delay (by up to 15%), and maintaining load balancing (with a 42% lower standard deviation in controller load). For future work, we aim to improve trajectory prediction accuracy using advanced deep learning techniques, adopt adaptive machine learning algorithms for controller placement, and address key challenges such as scalability for city-wide deployments and integration with 5G/6G network infrastructures. 7.1 Limitations of the Proposed Approach Although our proposed approach MOTA-SVB significantly improves network performance, it still has some limitations which will be discussed next. Addressing these limitations is considered as future work. 1. MOTA-SVB assumes an SDN architecture in vehicular networks where controllers are deployed at RSUs and vehicles act as data plane devices with mobility across the network. The communication infrastructure is assumed to be fully deployed, providing comprehensive coverage via RSUs, and all RSUs possess adequate computational capacity to host SDN controllers. This is not always the case and for such a scenario, one will need a different approach to improve the network performance. 2. MOTA-SVB does not include by default any security aspects (e.g. encryption, authentication, and integrity mechanisms) for vehicle-to-controller or inter-controller communications. To prevent vulnerability to eavesdropping, spoofing, or man-in-the-middle attacks, existing security approaches such as [64–66] can be integrated with our proposed approach. Closer integration with security solutions is also mentioned as a future research direction to achieve secure data transmission along with performance optimization. 3. MOTA-SVB does not explicitly simulate Byzantine behaviors (e.g., inconsistent controller responses, malicious controllers, or delayed and incorrect flow rule installations), as MOTA-SVB focuses on BFT-aware SDN controller placement rather than Byzantine behavior occurrence. Therefore, explicitly modeling and evaluating Byzantine behaviors will be considered as an important future research direction, enabling to evaluate the effectiveness of the MOTA-SVB approach under adversarial conditions. 4.

    MOTA-SVB: A Novel Multi-Objective Trajectory-Aware SDN Controller Placement in Vehicular Networks for Byzantine Fault Tolerance · 2026 · DOI

  • Despite its promise, AI-augmented orchestration faces significant deployment challenges. The first challenge is data fragmentation. Many networks still expose heterogeneous schemas, incompatible telemetry streams, and inconsistent timing granularity. Without a unified data model, AI decisions may be based on incomplete or conflicting information [10, 11, 16]. The second challenge is trust and explainability. Operators are unlikely to fully delegate control of mission-critical services to opaque models. Practical systems therefore need interpretable outputs, confidence estimates, safe exploration boundaries, and rollback mechanisms. The third challenge is interoperability. Cross-domain orchestration depends on open interfaces among RAN vendors, cloud platforms, transport controllers, and application orchestrators [6, 13, 14, 19]. In many real deployments, such openness is incomplete. The fourth challenge is security. An orchestration platform becomes a high-value control inputs, poisoned training It must be protected against policy tampering, adversarial point. data, and privilege escalation. The fifth challenge is organizational alignment. Even when the technology is available, different operational teams often optimize different KPIs. Cross-domain orchestration therefore requires not only new software, but also new workflows and governance structures.

    AI-Augmented Cross-Domain Resource Orchestration in Next-Generation Mobile Networks · 2026 · DOI

  • Lack of visibility into underlying physical networks in SD-WANs. Inability to make underlay-aware decisions in SD-WANs.

    Seeing Through the Overlay: Unveiling Potential Bottlenecks in SD-WANs · 2026 · DOI

  • The need for secure IT/OT convergence and edge-driven automation in industrial environments - The limitations of traditional security models and hub-and-spoke WAN architectures

    SASE-Based Enterprise Architecture Modernization Through Secure IT/OT Convergence and Edge-Driven Automation in Industrial Environments · 2026 · DOI

  • Recovery exercises conducted annually — the standard practice in traditional DR programs — are insufficient to maintain operational confidence in infrastructure and application cloud DR environments where configurations change continuously.

    SASE-Based Enterprise Architecture Modernization Through Secure IT/OT Convergence and Edge-Driven Automation in Industrial Environments · 2026 · DOI

  • The paper identifies a gap in the existing research on hexagonal architecture for maintainable applications. It highlights the need for a holistic perspective that integrates technical analyses with organizational and practical considerations.

    Hexagonal Architecture for Maintainable Applications · 2026 · DOI

  • The gap in existing research is the need for a holistic perspective that integrates technical analyses with organizational and practical considerations. The study identifies the need for more secure systems that incorporate cybersecurity measures from the inception of the design.

    Serverless Architecture: Patterns and Pitfalls · 2026 · DOI

  • The paper identifies a gap in the development of robust defense frameworks for machine learning systems. The study highlights the need for a holistic perspective on the design, implementation, and evaluation of robust solutions for contemporary security and architectural challenges. The paper addresses the lack of integration of cybersecurity measures with architectural design.

    Space-Based Architecture for Distributed Systems · 2026 · DOI

  • Static pathfinding algorithms have limitations in large-scale OHT systems. Existing RL approaches suffer from scalability issues.

    Lightweight Probabilistic RL for Web-app Compatible Large-scale OHT Path Optimization · 2026 · DOI

  • The paper identifies the limitations of 5G access networks, such as vendor lock-in, high deployment costs, and limited upgrade flexibility. The paper aims to address these challenges by proposing a unified resource management framework for a 6G-based network architecture.

    Energy-Efficient Resource Management via Hierarchical Reinforcement Learning in O-RAN · 2026 · DOI

  • The proposed framework is simulated, and the results may not generalize to real-world scenarios. The framework is evaluated using a single agent and a single learner. The framework does not address the challenge of scaling to multiple agents and learners.

    Toward Optimizing Reinforcement Learning Workload Placement at the Cloud-Edge Continuum in 6G Networks: A Scaled RL Framework · 2026 · DOI

  • The placement of RL workloads in the cloud-edge continuum is a significant challenge. Existing solutions typically exclude RL techniques due to their distinct structure and operational requirements. There is a need for a framework that enables the scaling of RL actor processes across both domains.

    Toward Optimizing Reinforcement Learning Workload Placement at the Cloud-Edge Continuum in 6G Networks: A Scaled RL Framework · 2026 · DOI

  • Evaluating the proposed policy in real-world deployments. Investigating the application of the policy in other domains requiring federated learning and real-time visualization.

    SLA-Driven Adaptive FL policy with Real-Time Visualization for Zero-Touch 6G Network Slicing · 2026 · DOI

  • The lack of scalable and sustainable AI-driven zero-touch automation in 6G networks. The limitations of traditional centralized approaches in terms of communication overhead and latency.

    SLA-Driven Adaptive FL policy with Real-Time Visualization for Zero-Touch 6G Network Slicing · 2026 · DOI

  • Evaluating SliceScope in a large-scale network. Investigating the performance of SliceScope in scenarios with high packet loss or latency. Extending SliceScope to support multiple network domains and slices.

    Dynamic SLA-aware Network Slice Monitoring · 2026 · DOI

  • Existing solutions lack end-to-end visibility or control mechanisms to dynamically allocate monitoring resources. There is a need for a system that can provide accurate end-to-end tracking of per-slice SLA metrics.

    Dynamic SLA-aware Network Slice Monitoring · 2026 · DOI

  • The need for deterministic memory management in highly resource-constrained edge hardware. The lack of a state-of-the-art framework for untrusted, massive-scale federated learning and autonomous spatial mapping.

    Optimizing Sovereign Edge Processing and Byzantine Fault Tolerance · 2026 · DOI

  • Existing QoS prediction approaches often rely on static, reactive models. These models fail to capture temporal traffic dynamics and struggle with class imbalance.

    A multi-model deep learning approach for proactive QoS prediction in 5G network slicing · 2026 · DOI

  • Lacks empirical validation; does not address hybrid architectures, imbalance, or operational constraints No practical implementation; lacks dynamic, closed-loop QoS prediction models Relies on simulation; singletask prediction; no imbalance or hybrid modeling High false positives; lacks temporal modeling and imbalance awareness Synthetic-only validation; no integration with QoS prediction pipelines Focuses on classification only; ignores continuous QoS prediction and temporal…

    A multi-model deep learning approach for proactive QoS prediction in 5G network slicing · 2026 · DOI

Most-cited papers in Software-Defined Networks and 5G

Most recent work

Find a gap in your own Software-Defined Networks and 5G sub-topic

This page shows what the Software-Defined Networks and 5G literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Computer Science

Online Learning and Analytics
851 open questions
Digital literacy in education
518 open questions
AI in Service Interactions
517 open questions
Music Technology and Sound Studies
423 open questions
Teaching and Learning Programming
384 open questions
Quantum Computing Algorithms and Architecture
368 open questions
Virtual Reality Applications and Impacts
317 open questions
Intelligent Tutoring Systems and Adaptive Learning
298 open questions
56 open questions have been extracted from the limitations and future-work passages of 204 Software-Defined Networks and 5G papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.