How to use this list
Public-key cryptography in everyday use rests on problems such as factoring and discrete logarithms. Shor (1997) showed that a sufficiently large quantum computer could solve both in polynomial time, which would break widely used schemes such as RSA. That is why organisations are preparing to migrate to post-quantum cryptography, algorithms that run on ordinary computers but are designed to resist known quantum attacks. Standards bodies, including the US National Institute of Standards and Technology, have begun publishing such algorithms, so the research question has shifted from "which algorithm?" to "how do we migrate safely?".
The four topics below each focus on one problem. Each row gives a possible research question, a likely design and why it matters now. For broader security topics see cybersecurity topics, cybersecurity in business and blockchain and IoT security.
Keep three terms apart when you write your proposal. Post-quantum cryptography means new mathematical algorithms for key exchange and signatures that run on existing hardware. Quantum key distribution uses quantum physics to share keys and needs dedicated optical equipment, so it suits only some links. AI-based defence uses machine learning to detect attacks; it can support either approach but provides no cryptographic protection of its own.

Theme 1: Post-quantum cryptography on constrained devices
| Topic | Possible research question | Likely design and data | Why it matters now |
|---|---|---|---|
| Post-quantum key exchange and signatures on constrained IoT devices | What are the latency, memory, energy and bandwidth costs of standardised post-quantum algorithms on typical IoT microcontrollers, and which protocol settings keep them within device limits? | Laboratory benchmarking on two or three hardware classes; repeated measurements; comparison with classical baselines | Combining AI-based and post-quantum approaches for IoT is an active review topic (Saeed & Alqahtani, 2025) |
Evidence that the gap is real. Saeed and Alqahtani (2025) review how AI-based and post-quantum approaches are being applied to IoT security. Because IoT devices have tight memory, power and bandwidth budgets, measured performance on real hardware is more useful to practitioners than further framework proposals. A benchmarking thesis also produces reusable data and code, which strengthens its contribution.
Theme 2: Migration in critical infrastructure and enterprise networks
| Topic | Possible research question | Likely design and data | Why it matters now |
|---|---|---|---|
| Planning the post-quantum migration in critical infrastructure | How do operators in one sector (for example energy or water) inventory their cryptography and decide what to migrate first, and what constraints slow them down? | Multiple case study; interviews with security and operations staff; analysis of migration plans | Operational technology has long lifecycles and safety and availability needs that limit change (Stouffer et al., 2023) |
| Hybrid classical and post-quantum key exchange in an enterprise network | What compatibility failures and performance costs arise when hybrid key exchange is enabled across an enterprise's existing services, and how can they be predicted? | Testbed experiment replicating a representative network; traffic and failure logs; cost model | Running classical and post-quantum algorithms together is a transition path that organisations need operational evidence on before adopting it |
Evidence that the gap is real. NIST guidance on operational technology security notes that industrial control systems often run for many years, prioritise availability and safety, and cannot be patched or reconfigured as freely as office IT (Stouffer et al., 2023). These constraints make cryptographic migration in critical infrastructure an organisational and engineering problem, not only a mathematical one, and a good fit for case-study or testbed research.
Theme 3: AI-assisted defence and its limits
| Topic | Possible research question | Likely design and data | Why it matters now |
|---|---|---|---|
| Robustness of machine-learning intrusion detection across IoT datasets | Do intrusion detection models trained on one IoT dataset keep their performance on another, and under adversarial manipulation of traffic features? | Cross-dataset experiments on public IoT datasets; adversarial evaluation following a recognised taxonomy; ablation studies | IoT intrusion datasets differ in features and attack labels, and ML models face known adversarial attacks (Booij et al., 2022; Vassilev et al., 2025) |
Evidence that the gap is real. Booij et al. (2022) show that public IoT intrusion datasets differ in their features and attack types, which makes results hard to compare; the widely used Bot-IoT dataset was itself built in a simulated testbed (Koroniotis et al., 2019). The NIST taxonomy of adversarial machine learning describes evasion, poisoning and other attacks that apply to security classifiers (Vassilev et al., 2025). A thesis that tests whether "intelligent" defences hold up across data and under attack is more useful than one that adds AI to a quantum framework without evaluation.
How to test a topic before you commit
Use the general method in how to choose a PhD topic and how to select a PhD topic, or our topic selection support, and screen each idea with the FINER criteria (Hulley et al., 2013). In this field, check three things: whether you can get the hardware or testbed you need; whether the algorithms and libraries you plan to use are stable enough to benchmark; and whether an industry partner will share information about real systems.
- Hardware: list the boards, servers or testbeds you need and confirm you can use them for the whole project.
- Baselines: decide which classical configuration you will compare against, so that costs are measured, not assumed.
- Data: for AI topics, name at least two public datasets and check their licences and labelling.
- Partners: for migration case studies, agree confidentiality terms early; operators may not share details otherwise.
Worked example (illustrative). A researcher begins with the live title on "AI-driven quantum cybersecurity architectures for IoT". She splits it: the cryptographic question becomes a benchmark of two post-quantum key exchange settings on three microcontroller boards, measuring handshake time and energy per connection; the AI strand is dropped. The thesis now has a clear method, measurable outcomes and a result that practitioners can use.
Common mistakes
- Treating post-quantum cryptography, quantum key distribution and AI as one research area.
- Claiming that AI provides quantum resistance.
- Describing quantum attacks on deployed systems as current fact rather than a future risk.
- Proposing an "integrated architecture" for every sector with no plan to test it.
- Reporting ML detection accuracy on one dataset as if it generalised.
Key takeaways
- Shor's algorithm is why widely used public-key schemes need replacing before large quantum computers exist (Shor, 1997).
- Measured performance of post-quantum algorithms on constrained devices is a practical doctoral contribution (Saeed & Alqahtani, 2025).
- Migration in critical infrastructure is shaped by operational constraints as much as by mathematics (Stouffer et al., 2023).
- AI-based defences must be tested across datasets and against adversarial attack (Booij et al., 2022; Vassilev et al., 2025).
- Screen every shortlisted topic for feasibility, novelty, ethics and relevance (Hulley et al., 2013).
