Time-lock puzzles (TLPs) are cryptographic mechanisms that defer access to information until a specified future time. But can they be made scalable, verifiable, and fair enough to meet the demands of real-world use?
Curious about our solution? Dr Aydin Abadi, Lecturer in Cybersecurity at our School of Computing, is leading the research uncovering its implications.
Contents:
- Introducing Delegated Time-Lock Puzzles
- The problem: TLPs aren't built for scale
- Our solution: Efficient Delegated Time-Lock Puzzles
- Real -world impact
- Key results: massive reductions in client and server workload
- Conclusion
Introducing Delegated Time-Lock Puzzles
Time-lock puzzles allow a message to be locked today and only revealed after a certain amount of time has passed.
This idea has wide-reaching applications, from delayed cryptocurrency payments and sealed-bid auctions to time-based access control and zero-knowledge proofs.
In our latest work, we present the concept of Delegated Time-Lock Puzzles (D-TLPs), a new, scalable framework that supports secure outsourcing of time-lock puzzle generation and solving, even when different clients and servers have vastly different computational capabilities.
The full paper, Scalable Time-Lock Puzzles, is presented at ACM AsiaCCS 2025 and is a collaboration between researchers at Newcastle University, University College London, and the University of Oxford.
The problem: TLPs aren't built for scale
Existing TLPs often assume a server has the computational power to solve a puzzle within a time interval, which is an assumption that breaks down quickly in practice.
Imagine two clients each create a TLP: one is meant to unlock in 24 hours, the other in 40 hours.
A server receiving both would need to solve them in parallel, performing 64 hours of computation.
Worse still, if hundreds or thousands of puzzles are submitted, the server’s workload becomes unrealistic.
Current solutions, like chained or batchable TLPs, don’t help when time intervals vary or clients are unrelated.
Specifically, Chained Time-Lock Puzzles (C-TLP) allow a client to encode multiple puzzles, each separated by a fixed time interval. The server must solve these puzzles sequentially—one after another—to obtain each message on time, rather than solving them all in parallel.
However, C-TLP has two key limitations:
- only a single client can generate the entire chain, preventing multiple clients from linking their puzzles
- the interval between puzzles is fixed, making it difficult to support arbitrary delays without incurring significant computational and communication overhead
Batchable Time-Lock Puzzles, in contrast, allow multiple puzzles of different clients to be merged into a single composite puzzle. Solving this composite puzzle reveals all embedded messages at once. However, this approach is only suitable when all messages are intended to be disclosed simultaneously.
Our solution: Efficient Delegated Time-Lock Puzzles (ED-TLP)
We developed ED-TLP, the first protocol to allow clients and servers to delegate their TLP tasks to potentially untrusted third-party solvers. The protocol is modular, secure, and efficient, supporting:
- end-to-end delegation
- multiple puzzles with varying time intervals
- real-time verification of solutions using lightweight mechanisms
- fair payment via smart contracts, ensuring helpers only get paid if they deliver correct results on time
- upper time bounds, offering not just when a solution might be available, but when it must be.
What does this mean for our previous example?
The two puzzles can be securely combined into a chain by the two clients.
Just like before, the first puzzle takes 24 hours, but the second message can be obtained 16 hours later. Therefore, it retains the overall 40 hours delay, but imposes only 40 hours of total computation, rather than 64 hours. The more puzzles that are combined, the larger the efficiency savings. To illustrate how our protocol helps avoid excessive computation and retrieve messages with no more delay than intended, we have prepared a short animation:
Fig.1: The video compares the working of our new protocol against existing solutions. Initially, it shows three setters producing puzzles and sending them to a solver to solve in parallel; it extracts the messages from the puzzles on time, but uses a lot of resources. Worse scenarios follow with the three puzzles sent to a server with a only a single free CPU that is not as powerful as needed, so the puzzles need to be solved sequentially and individually slower than intended, resulting in overall much higher delays than the puzzle setters desired. Finally, it shows our protocol combining puzzles into a single puzzle that can be solved in as much time as the longest puzzle previously but on a single CPU. Furthermore, it shows how our protocol allows a weak server to securely delegate puzzle solving to a powerful helper through a smart contract that pays the helper when the puzzle solution is delivered. This video was created by co-author Dan Ristea.
Real world impact
TLPs have powerful real-world applications, especially when deployed at scale. Our scalable ED-TLP framework makes it feasible to support the following scenarios:
- secure information release in high-risk environments: In politically sensitive or high-risk contexts, individuals such as journalists, whistleblowers, and human rights activists often face serious threats for exposing sensitive truths. Our scalable time-lock puzzles can serve as a powerful protective tool: they allow users to pre-schedule the gradual release of information, ensuring that critical evidence is made public even if they are detained, silenced, or harmed. By eliminating the need to stay online or rely on a trusted intermediary, this approach can help deter violence, shift risk away from the individual, and guarantee that the message survives, even if the messenger does not.
- online education and examinations: For students with unreliable internet connections, our scalable TLPs can allow advance downloading of encrypted exam materials, which become accessible only at the scheduled start time. This approach will preserve exam integrity while ensuring access.
- scheduled financial transactions: Businesses and individuals can schedule sensitive financial operations, such as payments or investment allocations, without having to reveal transactional details to financial institutions beforehand. This mitigates risks of insider threats, where employees might leak or misuse privileged information much earlier than the scheduled time.
In each of these settings, our protocol ensures scalability and fairness, supporting thousands of independently timed puzzles while offloading computational work to helpers in a verifiable, privacy-preserving way.
Key results: massive reductions in client and server workload
We implemented ED-TLP and tested it on up to 10,000 puzzles. Results include:
- 99% reduction in client-side computation
- 100% reduction in server-side solving
- smart contract cost as low as 0.2 cents per puzzle
- low overhead, even for large-scale deployments
The source code is open-source and freely available.
Conclusion
Our work on scalable time-lock puzzles is a step toward making time-dependent cryptography more usable, efficient, and fair. By supporting secure delegation and varied timing intervals, our solution makes time-lock puzzles viable for a wide range of practical, large-scale applications, from finance and education to human rights and secure communications.
We are excited about the broader applications of this work and invite collaboration with researchers and practitioners in security, cryptography, and blockchain systems.
You might also like
- read the study: Scalable Time-Lock Puzzles
- learn more about Dr Aydin Abadi, Lecturer in Cybersecurity
- find out about Dr Abadi’s work in Privacy-preserving deduplication to enhance federated learning
- discover more research from our School of Computing
