15. Doctoral Research and Emerging IoT–Wireless Frontiers

Introduction and Context

Internet of Things (IoT) and wireless systems have developed from connected sensing and identification technologies into distributed cyber-physical infrastructures. Contemporary IoT systems combine sensing, computation, communication, artificial intelligence, actuation, cloud and edge platforms, and increasingly autonomous decision-making. Wireless networks are no longer merely communication pipes. They are becoming programmable, intelligent, sensing-capable, energy-aware, and integrated with physical environments.

This transformation creates an important doctoral research problem. The central question is not simply how to connect more devices, but how to formulate defensible knowledge about systems whose technical, social, physical, and organizational dimensions interact. A doctoral contribution may consist of a new protocol, architecture, algorithm, formal model, dataset, experimental method, evaluation framework, or theory of deployment. Its quality depends not only on novelty, but also on a clearly bounded research question, appropriate evidence, transparent methods, comparison with credible baselines, and reproducibility.

Several frontiers are particularly significant. Sixth-generation (6G) research seeks networks that support extreme connectivity, integrated sensing and communication, distributed intelligence, and new service models. Semantic communications question whether networks should transmit all data or only the information necessary for a task. Ambient and batteryless IoT challenge conventional assumptions about persistent power. Quantum-safe systems address the long-term security of constrained devices and wireless infrastructures. Autonomous cyber-physical ecosystems combine machine learning with physical control, creating both new capabilities and new safety risks.

This chapter presents a research-oriented framework for investigating these frontiers. It links research paradigms, evidence strategies, design methods, experimental practice, formal reasoning, causal inference, and reproducibility. The aim is to help doctoral researchers move from an attractive technology theme to an original, testable, and responsible contribution.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Distinguish exploratory, descriptive, explanatory, predictive, evaluative, and design-oriented IoT research questions.

2. Select and justify systematic review, design science, experimentation, simulation, testbed, formal-methods, causal-inference, and mixed-method approaches.

3. Design reproducible evaluations using credible baselines, appropriate metrics, uncertainty analysis, and transparent reporting.

4. Critically assess 6G, integrated sensing and communication, semantic communications, AI-native networks, ambient IoT, quantum-safe systems, and autonomous cyber-physical ecosystems.

5. Identify technical, ethical, governance, safety, privacy, and security risks in emerging wireless systems.

6. Formulate a doctoral research agenda that produces a meaningful and defensible contribution.

From Technology Theme to Researchable Problem

A technology theme is not yet a research problem. “AI for 6G” is a topic; “How does federated reinforcement learning affect reliability and energy consumption in multi-cell scheduling under non-stationary traffic?” is a researchable question. The latter identifies an intervention, context, outcomes, and conditions.

A useful problem statement contains five elements: a phenomenon, a deficiency in existing knowledge or practice, a population or system context, a proposed means of investigation, and the significance of the result. The gap may be a performance gap, such as inadequate latency under mobility; a knowledge gap, such as uncertainty about causal factors; a methodological gap, such as reliance on unrealistic simulation assumptions; or a deployment gap, such as the absence of evidence across heterogeneous devices.

Doctoral research should distinguish novelty from value. A minor variation of an existing optimization algorithm may be novel in implementation but weak in contribution. Conversely, a carefully designed replication that reveals the limits of a widely accepted result can be highly valuable. Contributions commonly take the following forms:

• Construct contribution: a new concept, metric, model, dataset, or taxonomy.
• Method contribution: a new protocol, algorithm, architecture, or research procedure.
• Empirical contribution: evidence from experiments, field deployments, or comparative studies.
• Theoretical contribution: an explanation, theorem, bound, or formal relationship.
• Design contribution: an artifact that solves a significant problem under explicit constraints.
• Socio-technical contribution: evidence about adoption, governance, organizational effects, or human participation.

A research question should be matched to a claim. A descriptive claim requires measurement; a causal claim requires a credible counterfactual; a performance claim requires comparison; a safety claim requires hazard analysis or formal evidence; and a generalization claim requires variation across contexts.

Research Paradigms and Evidence Strategies

Systematic and Scoping Reviews

A systematic review uses a planned protocol to identify, select, appraise, and synthesize relevant studies. The protocol should define databases, search strings, inclusion and exclusion criteria, quality assessment, coding categories, and synthesis procedures before interpretation begins. In IoT, systematic reviews are valuable for mapping protocols, energy-harvesting techniques, security vulnerabilities, or evaluation practices. A scoping review is more suitable when concepts are diverse, terminology is unstable, or the objective is to map an emerging field rather than estimate a single effect.

A doctoral review should not become an annotated bibliography. It should reveal patterns, contradictions, neglected contexts, methodological weaknesses, and opportunities for theory or design. Bibliometric analysis can expose collaboration and citation structures, but it does not substitute for technical appraisal. Quality assessment should consider whether a study defines its environment, reports parameter settings, uses realistic workloads, compares against meaningful baselines, and makes data or code available.

Design Science Research

Design science is appropriate when the research seeks to create and evaluate an artifact. The artifact may be an architecture, protocol, algorithm, hardware-software platform, dataset, or decision framework. A disciplined design science process identifies the problem, derives objectives from requirements, develops the artifact, demonstrates its use, evaluates utility and limitations, and communicates the contribution.

The artifact must be evaluated against the problem rather than against an abstract ideal. For example, a batteryless sensor protocol should be assessed not only for throughput but also for startup reliability, energy-neutral operation, data loss, and performance under intermittent power. Design decisions should be traceable to requirements, and evaluation should include conditions under which the design fails.

Experiments, Testbeds, and Field Studies

Controlled experiments are useful when variables can be manipulated and outcomes measured. A wireless experiment should specify channel conditions, topology, traffic, device hardware, firmware, interference, mobility, repetitions, random seeds, and stopping rules. A testbed provides realism and repeatability, but its scale, hardware diversity, and environmental representativeness must be stated.

Field studies capture operational complexity that laboratory experiments miss. They also introduce confounders such as weather, user behavior, maintenance schedules, and infrastructure failures. Combining laboratory and field evidence is often stronger than treating either as universally sufficient. Researchers should report both average performance and distributional behavior, including tail latency, outage probability, fairness, and failure recovery.

Simulation and Digital Twins

Simulation enables exploration of scale, rare events, and controlled counterfactuals. It is essential when experiments are expensive or unsafe. However, simulation results are conditional on assumptions. A wireless simulator may abstract propagation, queues, hardware energy consumption, or adversarial behavior. Validation should compare important simulated outputs with measurements, and sensitivity analysis should test how conclusions change when assumptions vary.

Digital twins extend simulation by maintaining a relationship with a physical asset or environment through data synchronization. Their research value depends on fidelity, latency, model updating, uncertainty representation, and governance of the underlying data. A visually impressive twin is not necessarily a valid scientific instrument.

Formal Methods and Safety Cases

Formal methods provide mathematical evidence about properties such as reachability, invariance, protocol correctness, timing, or absence of specified classes of failure. Model checking can explore finite-state representations, while theorem proving can establish general propositions. Runtime verification monitors whether an executing system satisfies formalized properties.

Formal analysis is especially relevant to autonomous cyber-physical systems. A safety case can connect claims, arguments, evidence, assumptions, and residual risks. Formal proofs do not guarantee real-world safety if the model omits relevant physics, faulty sensors, human intervention, or distribution shifts. Therefore, formal and empirical methods should be treated as complementary.

Causal Inference and Mixed Methods

Prediction and causation are different. A model may predict network congestion without showing that a particular scheduling policy caused an improvement. Causal research requires a defined intervention, outcome, treatment assignment mechanism, and plausible counterfactual. Randomized experiments are powerful where feasible. In operational networks, researchers may use quasi-experimental methods, difference-in-differences, instrumental variables, regression discontinuity, or causal graphs, provided their assumptions are defended.

Mixed methods are useful when technical performance and human or organizational consequences must be studied together. A sequential design might begin with qualitative interviews to identify deployment barriers, develop a protocol, and then evaluate it experimentally. An explanatory design might first detect an unexpected performance pattern and then use interviews or trace analysis to explain it. Integration must occur in the research question and interpretation, not merely by placing qualitative and quantitative results in separate chapters.

Reproducibility as a Research Capability

Reproducibility is more than publishing source code. It involves sufficient information for another researcher to reconstruct the computational and experimental conditions that generated a result. A strong IoT research package may include source code, firmware, configuration files, hardware specifications, datasets or data-generation procedures, container definitions, random seeds, analysis scripts, protocol versions, and documentation of unavailable dependencies.

Researchers should distinguish computational reproducibility, experimental reproducibility, and inferential reproducibility. A result may be regenerated from code but fail on different hardware; alternatively, independent researchers may obtain the same conclusion using a different implementation. Reproducible reporting should expose negative results, failed runs, excluded data, calibration procedures, and deviations from the original plan.

Evaluation should use baselines that are correctly implemented and tuned. Reporting only the best run is misleading. Means should be accompanied by variance or confidence intervals where appropriate. Statistical significance does not establish practical significance, and multiple comparisons can produce apparently strong findings by chance. Energy, latency, reliability, privacy, and computational cost should be considered jointly rather than optimized in isolation.

Emerging Wireless and IoT Frontiers

6G as a Research Program

6G is best understood as a research program rather than a single technology. Its anticipated directions include extremely dense connectivity, higher-frequency operation, non-terrestrial integration, programmable radio environments, distributed artificial intelligence, sensing-communication integration, and stringent requirements for reliability and energy efficiency. The research challenge is to determine which combinations are technically feasible, economically justified, and socially acceptable.

Claims about 6G should therefore be decomposed. A proposed architecture may be assessed through spectral efficiency, energy per delivered bit, end-to-end latency, reliability, positioning accuracy, resilience, and lifecycle cost. Researchers should avoid treating a high peak data rate as evidence of overall superiority. Propagation loss, blockage, hardware impairment, synchronization, backhaul, and deployment density can dominate practical performance.

Integrated Sensing and Communication

Integrated sensing and communication (ISAC) uses shared signals, spectrum, hardware, or processing for data transmission and environmental sensing. A system may infer range, velocity, angle, occupancy, or motion while simultaneously serving communication users. The research problem involves waveform design, beamforming, resource allocation, sensing-communication trade-offs, and privacy.

Evaluation requires task-specific sensing metrics rather than only communication metrics. Detection probability, false-alarm rate, localization error, resolution, and tracking stability should be related to throughput, reliability, and power. A critical issue is identifiability: different physical scenes may produce similar observations. Machine learning can improve inference but may conceal uncertainty and fail under domain shift.

Semantic Communications

Conventional communication seeks accurate delivery of symbols or packets. Semantic communication seeks delivery of information relevant to a task, meaning, or goal. A surveillance system may not need every video pixel if its objective is detecting a safety event. This approach can reduce traffic, but it introduces application-dependent definitions of meaning and error.

Semantic systems require explicit task objectives, shared context models, and measures of semantic fidelity. A semantic distortion metric should be connected to the consequences of incorrect decisions. Compression that preserves average classification accuracy may still be unacceptable if it removes rare safety-critical events. Researchers must also examine interoperability, bias in knowledge models, explainability, and the possibility that sender and receiver interpret context differently.

AI-Native Networks

AI-native networks use learning throughout planning, orchestration, resource management, protocol adaptation, and service assurance. Reinforcement learning may select channels or schedules; graph learning may model topology; federated learning may reduce raw-data movement; and generative models may assist configuration or diagnosis.

The principal research issue is not whether AI can produce a high score in a static simulation. It is whether learning systems remain safe, stable, efficient, interpretable, and adaptable under non-stationarity. Relevant risks include poisoned updates, adversarial examples, reward hacking, model drift, privacy leakage, unstable control loops, and unequal service quality. AI should be assigned bounded authority, monitored continuously, and supported by fallback policies and human override where consequences are significant.

Ambient and Batteryless IoT

Ambient IoT devices obtain energy from radio frequency fields, light, thermal gradients, vibration, or other environmental sources. Batteryless operation can reduce maintenance and environmental burden, but it produces intermittent availability. Communication and computation must be organized around energy opportunities rather than assumed continuous power.

Research questions include energy-neutral scheduling, wake-up detection, nonvolatile state, backscatter communication, intermittent computing, and data prioritization. Evaluation should report cold-start behavior, energy budget, capacitor or storage dynamics, packet loss, time to useful measurement, and performance under realistic harvesting variability. A batteryless device that performs well under constant laboratory illumination may fail in a warehouse or agricultural environment.

Quantum-Safe IoT and Wireless Security

Quantum-safe security concerns cryptographic mechanisms intended to resist attacks from quantum-capable adversaries. IoT research must consider algorithmic security alongside key size, message overhead, memory, computation, radio airtime, firmware update mechanisms, and device lifetime. Migration is difficult because many devices are constrained, widely deployed, and difficult to replace.

A sound research agenda includes crypto-agility, hybrid migration, secure bootstrapping, certificate management, inventory discovery, and long-term confidentiality. Security claims should specify the adversary, protected assets, trust assumptions, and failure recovery. Replacing one algorithm without redesigning provisioning, update, and revocation processes does not create a quantum-safe ecosystem.

Autonomous Cyber-Physical Ecosystems

Autonomous cyber-physical ecosystems contain multiple interacting agents, sensors, actuators, services, and human stakeholders. Examples include intelligent transport, collaborative robots, smart grids, and adaptive industrial systems. Their complexity arises from feedback, emergence, partial observability, competing objectives, and cascading failures.

Research must address safety envelopes, verification of learned components, human-machine authority, resilience to communication loss, and responsibility for decisions. A controller that is optimal in simulation may be unsafe when sensing is delayed or another agent behaves unexpectedly. Hierarchical autonomy, runtime assurance, conservative fallback controllers, and explicit operating boundaries are important design strategies.

A Doctoral Research Workflow

A practical workflow begins with a systematic mapping of literature and stakeholder needs. The researcher then defines the system boundary, research questions, hypotheses or design objectives, assumptions, and evaluation criteria. Next comes artifact or experiment design, followed by pilot testing. Pilot studies frequently reveal measurement errors, unrealistic workloads, synchronization problems, or hidden confounders.

The main study should use an evaluation matrix connecting each claim to evidence. For example, a claim of energy efficiency requires a power measurement protocol; a claim of robustness requires perturbation conditions; a claim of generality requires multiple environments; and a claim of safety requires hazards, constraints, and assurance evidence. Analysis should distinguish exploratory findings from confirmatory tests.

Finally, the researcher should explain boundary conditions. Every result has a domain of validity defined by hardware, workload, channel, topology, data distribution, threat model, and governance context. Stating these boundaries strengthens rather than weakens a dissertation because it prevents overgeneralization.

Illustrative Case Study: Batteryless Semantic Monitoring in a Smart Warehouse

Background

A logistics organization wishes to monitor package handling and detect dangerous impacts using small batteryless wireless tags. Replacing batteries at scale is expensive, while continuous transmission would exceed the available harvested energy. The proposed system uses backscatter communication, intermittent computing, and semantic reporting.

Situation and Problem

A conventional design transmits periodic accelerometer samples to an edge server. In preliminary trials, tags frequently lose power before completing a packet, and the network becomes congested during loading periods. The research question is whether a semantic protocol can preserve safety-relevant detection while reducing energy and airtime.

Analysis

The research team defines the semantic task as identifying impact events above a calibrated risk threshold, not reconstructing the entire acceleration waveform. However, rare events are safety-critical, so average classification accuracy is insufficient. The evaluation therefore includes missed-event rate, false alarms, time to alert, energy per decision, startup success, and performance across lighting and radio conditions.

A causal experiment varies reporting policy: raw periodic transmission, locally extracted features, and event-triggered semantic reporting. Warehouse zones, tag placement, package type, and illumination are treated as contextual variables. Randomized trial periods reduce confounding from time-of-day traffic. A digital model is used to explore larger tag populations, but conclusions are anchored in measured hardware data.

Approach and Solution

Each tag performs a low-power threshold test and stores a compact event summary in nonvolatile memory. When energy is insufficient for immediate transmission, the summary is retained until the next harvesting opportunity. The edge service combines the summary with contextual information and requests raw samples only when uncertainty is high. A safety rule prevents semantic compression from suppressing events whose amplitude exceeds a conservative threshold.

The protocol is evaluated against the raw-transmission baseline using repeated runs across representative zones. The research package includes firmware, configuration files, measurement scripts, calibration procedures, and anonymized traces. A failure analysis records missed wake-ups, corrupted state, radio blockage, and model uncertainty.

Outcome

The illustrative study would be considered successful only if it demonstrates a meaningful reduction in energy or airtime without unacceptable growth in missed safety events. If semantic reporting performs well in one zone but fails under metal obstruction, the appropriate conclusion is conditional effectiveness, not universal superiority. The result may still constitute a doctoral contribution through the protocol, evaluation method, and explanation of boundary conditions.

Learning Points

The case illustrates that emerging IoT research requires alignment among application meaning, energy behavior, communication design, statistical evaluation, and safety constraints. It also shows why a single metric such as throughput cannot establish system value.

Challenges, Ethics, and Governance

Emerging IoT research faces technical and epistemic risks. Simulations may hide implementation costs; benchmark datasets may not represent deployment populations; AI models may exploit artifacts; and complex architectures may make responsibility unclear. Common mistakes include changing baselines after seeing results, tuning a proposed method more carefully than competitors, reporting only averages, confusing correlation with causation, and treating proprietary testbed data as automatically valid.

Privacy is particularly important in sensing systems. ISAC, semantic monitoring, and autonomous environments can infer occupancy, behavior, location, or identity even when the system was designed for another purpose. Data minimization, purpose limitation, access control, retention limits, secure processing, and privacy-preserving analytics should be designed from the beginning. Consent may be inadequate where people cannot realistically avoid instrumented environments.

Governance should address accountability, auditability, model updates, incident response, procurement, and decommissioning. Safety-critical systems require clear authority boundaries between automated agents and humans. Researchers should disclose conflicts of interest, security vulnerabilities responsibly, and limitations that could affect public or occupational safety. Environmental assessment should include device manufacture, deployment, maintenance, energy use, and disposal rather than focusing only on operational efficiency.

Best Practices for Doctoral Work

A strong doctoral project in IoT and wireless systems should:

• define a precise contribution and system boundary;
• connect every claim to an appropriate form of evidence;
• use realistic baselines and report parameter choices;
• combine analytical reasoning with measurement where feasible;
• quantify uncertainty, failure, and resource cost;
• release artifacts while protecting sensitive data;
• test robustness under distribution shift and adverse conditions;
• separate exploratory analysis from confirmatory evaluation;
• document assumptions, exclusions, and negative results;
• include ethical, privacy, safety, and lifecycle analysis.

The most valuable frontier research is neither technology hype nor narrow optimization. It explains what works, why it works, when it fails, and what consequences follow from deploying it.

Current Directions for Research

Near-term research is likely to emphasize joint communication-sensing-control design, energy-aware intelligence at the edge, interoperable non-terrestrial and terrestrial networks, secure learning across organizations, and automated assurance of adaptive systems. Semantic networking will require better task-specific metrics and shared context management. Batteryless devices will motivate new data models based on intermittent availability. Quantum-safe migration will shift attention from algorithms alone to lifecycle architecture and fleet management.

Across these areas, reproducibility and responsible experimentation will become competitive advantages. As systems become more autonomous, the ability to justify decisions, detect uncertainty, recover from failure, and demonstrate compliance will be as important as raw performance.

Chapter Summary

Doctoral research in IoT and wireless systems requires more than implementing a new algorithm. It requires a defensible relationship among problem, contribution, method, evidence, and boundary conditions. Systematic reviews identify knowledge structures and gaps; design science creates and evaluates artifacts; experiments, testbeds, simulation, and digital twins provide complementary evidence; formal methods support assurance; causal inference clarifies intervention effects; and mixed methods connect technical performance with human and organizational realities.

The chapter examined 6G, ISAC, semantic communications, AI-native networks, ambient and batteryless IoT, quantum-safe systems, and autonomous cyber-physical ecosystems. Each frontier introduces distinctive metrics, assumptions, and risks. Across all of them, reproducibility, realistic evaluation, uncertainty analysis, privacy, safety, governance, and lifecycle thinking are essential. A meaningful doctoral contribution explains not only improvement, but also mechanism, context, limitations, and consequences.

Key Terms

Ambient IoT: IoT systems powered by environmental energy or opportunistic wireless energy rather than conventional batteries.

Autonomous cyber-physical ecosystem: A network of interacting computational, physical, and human agents capable of adaptive decision-making.

Causal inference: Methods for estimating the effect of an intervention using counterfactual reasoning and explicit assumptions.

Design science: A research paradigm centered on constructing and evaluating artifacts that address identified problems.

Digital twin: A data-connected computational representation of a physical asset, process, or environment.

Formal methods: Mathematical techniques used to specify, verify, or prove properties of systems.

Integrated sensing and communication: Joint use of wireless resources for environmental sensing and data communication.

ISAC: Abbreviation for integrated sensing and communication.

6G: A prospective generation of wireless systems emphasizing integrated intelligence, sensing, extreme connectivity, and new network architectures.

Intermittent computing: Computation designed to continue or recover despite irregular energy availability.

Semantic communication: Communication optimized for task-relevant meaning rather than complete symbol-level transmission.

Quantum-safe security: Security designed to resist attacks by cryptanalytically relevant quantum computers.

Reproducibility: The ability to regenerate or independently obtain results through transparent methods and artifacts.

Runtime verification: Monitoring an executing system to determine whether specified properties are satisfied.

Semantic fidelity: The degree to which transmitted information preserves task-relevant meaning.

Testbed: A controlled physical or virtual environment for repeatable system evaluation.

Threat model: An explicit description of adversaries, capabilities, assets, and attack assumptions.

Uncertainty analysis: Examination of variability, measurement error, model assumptions, and confidence in results.

Review Questions

Short Answer Questions

1. Why is a technology theme insufficient as a doctoral research problem?

2. What is the difference between computational and experimental reproducibility?

3. Define integrated sensing and communication.

4. Why are average accuracy and throughput insufficient for evaluating safety-critical semantic communication?

5. What is the purpose of a threat model?

Long Answer Questions

1. Explain how systematic reviews, design science, experiments, simulation, and formal methods can be combined in an IoT doctoral study.

2. Discuss the principal research challenges of AI-native wireless networks.

3. Analyze the technical and methodological implications of ambient and batteryless IoT.

4. Explain why quantum-safe IoT requires more than selecting a post-quantum cryptographic algorithm.

5. Discuss how autonomous cyber-physical ecosystems should be evaluated for safety, resilience, and accountability.

Analytical / Application Questions

1. Design an evaluation matrix for a semantic communication protocol used in an industrial safety application. Identify claims, metrics, baselines, and evidence.

2. A wireless learning algorithm improves mean latency in simulation but performs poorly in a field trial. Propose a mixed-method investigation to explain the discrepancy.

3. Formulate a doctoral research question involving batteryless sensing and causal inference. Specify the intervention, outcome, confounders, and required evidence.

22. Multiple Choice Questions with Answers

Question 1

Which research question is most appropriately formulated as a doctoral IoT research problem?

A. What is 6G?

B. How can wireless technology be improved?

C. How does federated scheduling affect reliability and energy use under non-stationary traffic?

D. Why is IoT important?

Correct Answer: C. How does federated scheduling affect reliability and energy use under non-stationary traffic?

Question 2

A researcher wants to estimate the effect of a new scheduling policy on latency. Which concept is most directly required?

A. Counterfactual reasoning

B. Peak data rate

C. Bibliometric centrality

D. Semantic compression

Correct Answer: A. Counterfactual reasoning

Question 3

Which evidence is most appropriate for a claim that a protocol is safe under specified operating conditions?

A. A single average-throughput measurement

B. A formal safety argument supported by testing

C. A citation count

D. A larger training dataset alone

Correct Answer: B. A formal safety argument supported by testing

Question 4

Why should a batteryless IoT system report cold-start behavior?

A. Because batteryless devices always have higher bandwidth

B. Because operation may depend on intermittent energy availability

C. Because cold-start behavior measures spectral efficiency directly

D. Because it replaces the need for field testing

Correct Answer: B. Because operation may depend on intermittent energy availability

Question 5

What distinguishes semantic communication from conventional packet delivery?

A. It always transmits more raw data

B. It optimizes transmission for task-relevant meaning

C. It eliminates the need for shared context

D. It guarantees zero classification error

Correct Answer: B. It optimizes transmission for task-relevant meaning

Question 6

Which metric combination is most suitable for evaluating ISAC in a safety application?

A. Only peak download speed

B. Detection probability, false alarms, latency, and communication reliability

C. Only the number of neural-network parameters

D. Citation impact and software popularity

Correct Answer: B. Detection probability, false alarms, latency, and communication reliability

Question 7

A simulation result changes substantially when propagation assumptions are varied. What should the researcher do?

A. Report only the most favorable assumption

B. Remove the sensitivity analysis

C. Report the dependence and investigate validation with measurements

D. Declare the result universally valid

Correct Answer: C. Report the dependence and investigate validation with measurements

Question 8

Which statement best describes quantum-safe IoT migration?

A. It concerns only replacing encryption algorithms

B. It includes cryptography, provisioning, updates, key management, and constrained-device costs

C. It is unnecessary for long-lived devices

D. It eliminates the need for threat modeling

Correct Answer: B. It includes cryptography, provisioning, updates, key management, and constrained-device costs

Question 9

What is a major risk of unconstrained AI control in autonomous cyber-physical systems?

A. The system cannot use sensors

B. Reward hacking or model drift may produce unsafe actions

C. The system always becomes deterministic

D. Formal methods become unnecessary

Correct Answer: B. Reward hacking or model drift may produce unsafe actions

Question 10

Which practice most strongly supports reproducible IoT research?

A. Reporting only the best experimental run

B. Changing baselines after observing results

C. Releasing code, configurations, seeds, hardware details, and analysis procedures

D. Using proprietary data without documentation

Correct Answer: C. Releasing code, configurations, seeds, hardware details, and analysis procedures