Device independent quantum key distribution with robust self-tests

1. Problem Statement

Device-Independent Quantum Key Distribution (DIQKD) aims to generate secure keys based solely on the observed statistics of quantum measurements, without trusting the internal workings of the devices. While this offers high security against device imperfections, it suffers from severe practical limitations:

• Low Key Rates: DIQKD protocols typically require extremely high detection efficiencies to tolerate noise, often making them infeasible over long distances due to the "detection loophole."
• Optimization Complexity: Security proofs require optimizing over all possible quantum models compatible with observed statistics, a computationally intractable problem for general scenarios.
• The Gap: There is a theoretical gap between fully device-independent models and device-dependent models (where devices are trusted). In device-dependent settings, protocols like BB84 can achieve optimal key rates (Shor-Preskill bound), but these require trusting the hardware.

The paper addresses the challenge of bridging this gap. Specifically, it investigates "Routed Bell Test" setups, where local parties (Alice and Bob) perform a Bell test with nearby partners (Fred and George) to "self-test" their devices, while simultaneously performing a key-generation protocol (e.g., BB84) with each other over long distances. The goal is to rigorously prove that if local self-tests are performed, the security analysis can be "lifted" from a complex DIQKD optimization to a simpler, device-dependent optimization, even in the presence of experimental noise.

2. Methodology

The authors develop a rigorous mathematical framework combining Routed Bell Tests with Robust Self-Testing.

A. Modeling Routed Bell Tests

The paper compares two existing models for routed Bell tests:

1. Switch Model (Fig 1a): A source sends states to a switch, which routes them either to a local test partner (Fred) or the distant partner (Bob).
2. Source Model (Fig 1b): The source itself is conditioned on a classical variable $T$ to emit different states.

The authors prove that under a specific Marginal Constraint (the state received by Alice must be identical regardless of the switch setting $T$), these two models are statistically equivalent. They demonstrate via a counter-example (Example 2.1) that without this constraint, a malicious device could leak information about the round type (test vs. key generation), breaking security.

B. Robust Self-Testing Framework

The core methodology relies on the concept of Robust Self-Testing (from reference [36]).

• Exact Self-Testing: If a specific set of statistics (e.g., maximal CHSH violation) is observed, the underlying quantum state and measurements are uniquely determined (up to isometries).
• Robustness: In reality, statistics are imperfect. The authors utilize the property that if observed statistics are $\epsilon$-close to the ideal self-test statistics, the underlying physical state is close to the ideal state in terms of isometric dilation.
• The "Lift": The paper constructs a mathematical "lift" that maps the physical, unknown device-dependent state to an ideal, known state. This allows the security analysis to be performed on the ideal state with an error bound derived from the robustness of the self-test.

C. Entropy Transfer and Optimization

The authors establish a theorem (Theorem 3.9) showing that the conditional von Neumann entropy (which determines the key rate) in the device-independent setting can be lower-bounded by the entropy in the ideal, device-dependent setting, minus a term dependent on the self-test error ($\epsilon$).

• They define a compression of correlations (Lemma 3.7), showing that statistics from the physical setup can be approximated by statistics from the ideal model within a small error margin ($3(\epsilon_A + \epsilon_B)$).
• They prove Entropy Transfer (Lemma 3.8), showing that the uncertainty about the key (conditional entropy) in the physical system is close to that of the ideal system.

3. Key Contributions

1. Rigorous Equivalence of Models: The paper provides a formal proof that different routed Bell test models are equivalent under the marginal constraint, clarifying the necessary assumptions for security.
2. Robust Lift to Device-Dependent QKD: The primary contribution is the proof that a DIQKD protocol with local self-tests can be mathematically "lifted" to a device-dependent optimization problem. This transforms a hard, infinite-dimensional DIQKD optimization into a finite-dimensional, tractable problem.
3. Quantitative Error Bounds: The authors derive explicit bounds on the key rate degradation based on the quality of the local self-test. They show that if the local CHSH test is $\epsilon$-close to optimal, the key rate approaches the optimal Shor-Preskill rate with a penalty of order $O(\sqrt{\epsilon})$.
4. Finite-Size Analysis Framework: The paper outlines how to apply these theoretical results to finite-size scenarios using Renyi-entropy accumulation, providing a pathway for practical experimental implementation.

4. Key Results

• Theorem 3.5 & 3.9: These theorems establish that for a routed BB84 protocol, if Alice and Bob perform local CHSH tests with Fred and George respectively, the extractable key rate $r_{key}$ satisfies:
  $$r_{key} \geq 1 - h(Q_X) - h(Q_Z) - O(\sqrt{\epsilon})$$
  where $Q$ represents bit-error rates and $\epsilon$ represents the deviation from the ideal self-test. As $\epsilon \to 0$, the protocol achieves the optimal device-dependent BB84 rate.
• Marginal Constraint Robustness: The paper proves that the security is robust even if the marginal constraint (Alice's state independence from the switch) is only approximately satisfied, provided the trace distance deviation is small.
• Numerical Implication: The framework allows researchers to use existing, efficient numerical tools for device-dependent QKD (like semidefinite programming) to analyze DIQKD scenarios, provided local self-tests are included.

5. Significance

• Practicality for Long-Distance QKD: This work offers a realistic path toward long-distance DIQKD. By using local self-tests to certify devices, the protocol can tolerate the high losses associated with long-distance transmission (which usually kill standard DIQKD) while maintaining the security guarantees of device-independence.
• Bridging Theory and Experiment: It resolves the tension between the abstract, assumption-free nature of DIQKD and the concrete, high-performance nature of device-dependent QKD. It suggests that "routed" setups are not just theoretical curiosities but practical tools for next-generation quantum networks.
• Security Certification: It provides a method to certify that a device behaves as a specific quantum model (e.g., a qubit) without trusting the manufacturer, simply by observing local statistics. This is crucial for the commercialization of quantum cryptography.

In summary, the paper demonstrates that local self-testing acts as a bridge, allowing the high security of device-independence to be combined with the high efficiency of device-dependent protocols, provided the experimental setup includes a "routed" architecture with nearby test partners.