⚛️ quantum physics

Finite-Key Readiness and Authentication-Key Replenishment in QKD-Assisted UAV Edge Computing

This paper quantifies the performance cost of finite-key readiness and authentication-key replenishment in QKD-assisted UAV edge computing, demonstrating that while fresh keys significantly increase task completion time and energy consumption, a carefully optimized pulse budget can still meet strict deadlines under specific optical loss and misalignment conditions.

Original author: Dang H. Le, Ban Q. Tran, Susan Mengel

Published 2026-11-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original author: Dang H. Le, Ban Q. Tran, Susan Mengel

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

In the world of secure communication, there is a growing race to build systems that cannot be broken by future computers. Traditional security relies on complex math problems that are hard to solve today but might be easy for powerful quantum computers tomorrow. To counter this, scientists use a method called quantum key distribution. This technique creates secret codes based on the laws of physics rather than math. If someone tries to eavesdrop on the process, the laws of physics themselves change the message, alerting the users that the line is compromised. This method is incredibly secure, but it is also slow and finicky. It requires a special handshake to create a fresh key every time a new message needs protection, and if that handshake fails, the whole process stops.

Now, imagine taking this delicate technology and putting it on a drone hovering in the sky. Unmanned aerial vehicles are increasingly used to provide computing power to remote areas, acting as floating data centers. However, these drones burn a tremendous amount of energy just to stay in the air. Every second they hover while waiting for a secure key to be generated is a second of wasted fuel. Researchers wanted to know: if a drone has to wait for a fresh, physics-based secret key before it can do its job, how much time and energy does that cost? And more importantly, how much interference from the environment or a clever attacker can the system handle before it runs out of time or fuel?

A team of researchers set out to answer these questions by simulating a specific scenario. They imagined a trusted drone hovering in one spot, serving two different ground stations. The drone acts as a bridge, receiving data from the ground, processing it, and sending the results back. Before the drone can do any of this work, it must first establish a secure connection with each ground station using quantum keys. The researchers built a detailed computer model to track every single step of this process. They accounted for the time it takes to send the quantum signals, the time spent checking if the signals were received correctly, and the time required to process the data. Crucially, they also tracked the energy the drone consumed while waiting. They modeled a situation where the drone has a limited supply of pre-shared passwords used to verify the identity of the ground stations, and they watched to see if the drone would run out of these passwords while waiting for a successful connection.

The study focused on a realistic environment where the signal between the ground and the drone is weakened by distance and atmospheric conditions, a factor the researchers measured at a specific level of loss. They found that the requirement for a fresh key significantly changes the performance of the system. In their simulations, when the system had to wait for a new key to be generated and verified, the time it took to complete a task increased dramatically. Specifically, the time by which 95 percent of tasks were finished jumped from just over ten seconds to nearly twenty-nine seconds. This delay is not just a matter of speed; it directly translates to energy. Because the drone must keep its rotors spinning while it waits, the energy consumption grows in direct proportion to the delay. The researchers discovered that the system operates within a very narrow window of possibility. If the signal loss is too high, or if the ground stations are slightly misaligned, the drone cannot generate enough keys in time to meet its deadline, and the service fails.

The team also tested how the system would hold up against an active attacker trying to disrupt the connection. They simulated an adversary who could introduce small amounts of noise or extra signal loss to confuse the system. The results were stark: even a tiny increase in error or signal loss was enough to push the system past its deadline, causing the drone to fail its mission. Furthermore, these failed attempts were not free. Every time the drone tried to connect and failed, it burned through a portion of its limited supply of authentication passwords. If an attacker forced enough failures, the drone would eventually run out of passwords entirely, leaving it unable to communicate securely at all. The researchers noted that while the mathematical security of the keys themselves remained unbroken, the practical ability to use them was destroyed by the delay and the exhaustion of resources.

One of the most surprising findings was that the cost of verifying the identity of the messages was surprisingly low, but the cost of waiting for the keys was high. The researchers compared their quantum system to a newer, math-based security method that does not require waiting for a fresh key. While the math-based method is faster, it relies on the assumption that no one can solve the underlying math problem, an assumption that could be broken in the future. The quantum method offers a guarantee based on physics, but the price of that guarantee is time and energy. The study showed that for the drone to succeed, the signal quality must be exceptionally good. At a certain level of signal loss, the system simply cannot meet the deadline, no matter how much computing power is available. The researchers concluded that for this technology to work in real-world edge computing, the system must be designed with a massive margin for error, or the drone must be able to hover for much longer than is currently practical.

The work highlights a fundamental trade-off in the future of secure computing. We can have keys that are unbreakable by any computer, but getting them takes time, and time costs energy. For a drone hovering in the sky, that energy cost is the difference between completing a mission and running out of fuel. The study does not propose a new way to fix this, but rather provides a clear map of the problem. It shows exactly where the system breaks down, how much interference it can tolerate, and how much energy is wasted in the process. By quantifying these costs, the researchers provide a realistic picture of what it takes to bring quantum security to the skies. The findings suggest that while the technology is theoretically sound, its practical application in mobile, energy-constrained environments like drones requires careful management of signal quality and strict limits on what the system is expected to achieve. The window for success is small, and the margin for error is thin, but understanding these limits is the first step toward building systems that can operate safely in the real world.

Technical Summary: Finite-Key Readiness and Authentication-Key Replenishment in QKD-Assisted UAV Edge Computing

Problem Statement
The paper addresses the operational latency and resource constraints introduced when integrating Quantum Key Distribution (QKD) into Unmanned Aerial Vehicle (UAV) edge computing services. Unlike asymptotic QKD analyses, real-world implementations rely on finite-key protocols where a key is only released if a specific block of data passes statistical acceptance tests; otherwise, the attempt aborts. For a UAV serving ground gateways, this "fresh-key" policy (where keys are not pre-shared but generated on-demand) places the QKD attempt on the critical path of the service.

The core challenge is quantifying the cost of this policy in terms of time (deadline reliability) and energy (hover consumption), while simultaneously managing the depletion of authentication keys. Every QKD attempt consumes pre-shared authentication bits for message authentication (Wegman-Carter). If the protocol aborts frequently due to channel noise or finite-key fluctuations, the authentication pool may be exhausted before a valid key is generated, leading to a Denial of Service (DoS). The authors investigate the trade-offs between pulse budgets, optical loss, misalignment, and deadline constraints in a scenario where a stationary UAV serves two ground gateways sequentially.

Methodology
The study employs a failure-aware service model combining a count-level Monte Carlo simulation with a semi-analytic model.

System Topology: A trusted stationary UAV hosts a QKD receiver and a Mobile Edge Computing (MEC) server, serving two ground gateways (A and B) sequentially. Each gateway performs one QKD attempt followed by a protected upload, computation, and download.
QKD Protocol: The system uses a finite-key, decoy-state BB84 protocol with parameters fixed in advance. The protocol either outputs a key of predetermined length or aborts. The security parameter is set to $\epsilon = 2 \times 10^{-12}$ .
Authentication: Each attempt consumes 1,440 pre-shared authentication bits (18 pads and 2 selectors) with no recycling. The authentication scheme uses Wegman-Carter hashing.
Simulation & Modeling:
- Monte Carlo: Simulates optical detection statistics (multinomial/hypergeometric), acquisition outcomes, RF fading, and cryptographic processing costs. It tracks the "95th-percentile completion time" and the probability of meeting a deadline.
- Semi-Analytic Model: Reproduces the simulation results by assuming independent pulses and deterministic latency given acceptance, validating the approximations used in the simulation.
Threat Model: The analysis considers an external adversary capable of injecting errors and adding loss to the quantum channel or forging/delaying messages on the classical channel. The goal is to evaluate how these disturbances affect deadline reliability and authentication pool sustainability.

Key Contributions

Failure-Aware Service Model: The authors link fixed-length finite-key BB84 attempts directly to protected computation workflows. Unlike previous works that treat key rates as continuous utilities, this model accounts for the abort time and hover energy consumed by failed attempts. The semi-analytic model reproduces the simulated 95th-percentile completion time within 0.03 seconds.
Deadline–Replenishment Window: The study identifies a "window" for the pulse budget ( $N$ $N$ ) bounded by two constraints:
- Lower Bound (Replenishment): The pulse budget must be high enough to generate keys frequently enough to replenish the authentication pool before it is exhausted.
- Upper Bound (Deadline): The pulse budget must be low enough to ensure the total service time (QKD + computation + transfer) meets the deadline.
- The authors demonstrate that gross scalar key rates fail to capture these bounds, while net rates recover the floor but miss the ceiling.
Quantified Robustness under Disturbance: The paper quantifies how small increases in error rates or optical loss (e.g., 0.2 dB extra loss or a 0.24-point rise in error rate) can push the deadline probability well below the target (95%) and drain authentication keys, even if the protocol's theoretical security parameter remains unchanged.

Results

Latency Impact: At an effective optical loss of 26 dB, the requirement for fresh keys increases the 95th-percentile completion time for both tasks from 10.37 s to 28.65 s. Hover energy consumption grows proportionally with this time.
Authentication Cost: The "authentication debit" (bits burned per attempt) shifts the lower bound of the viable pulse budget. Hashing costs are shown to be near-zero relative to the deadline constraints when using proxy-benchmarked processing, but a "high-cost stress reference" (charging full hashing costs) narrows the viable window significantly.
Deadline Limits: At a 35-second deadline, the maximum tested optical loss allowing a non-empty pulse budget window is 27 dB (at 1% misalignment) and 22 dB (at 2% misalignment). At 3% misalignment, no tested setting satisfies the deadline.
Confidence: Selected configurations meet the 35-second deadline with simultaneous lower confidence bounds of 96.9–97.3% on held-out trials at 26 dB loss and 18–20 dB with 2% misalignment.
Vulnerability: The system is highly sensitive to disturbances. Small perturbations that preserve the theoretical security parameter can still defeat the availability goals (G3 and G4) by causing frequent aborts and draining the authentication pool.

Significance and Claims
The paper claims to provide a practical quantification of the "price" of information-theoretic security in a deadline-constrained, resource-constrained UAV environment. It moves beyond theoretical security proofs to evaluate availability and sustainability in the face of finite-key fluctuations and authentication key depletion.

The authors emphasize that while QKD offers security based on physics (unlike post-quantum cryptography which relies on computational hardness), this advantage comes with a tangible cost in time and energy. The study highlights that for low-latency services, the finite-key nature of the protocol creates a narrow operational window where key generation, authentication replenishment, and deadline adherence must be balanced. The work serves as a benchmark for understanding how much "margin" is required to maintain service reliability when QKD is placed on the critical path of edge computing tasks. The authors explicitly state they do not propose new security proofs or globally optimal schedulers, but rather characterize the trade-offs for fixed configurations under specific assumptions.

Technical Summary: Finite-Key Readiness and Authentication-Key Replenishment in QKD-Assisted UAV Edge Computing

More like this