One-sided DI-QKD secure against coherent attacks over long distances

This paper presents a one-sided device-independent QKD protocol that achieves security against coherent attacks with detection efficiencies above 50.1% on the untrusted side, enabling long-distance implementation comparable to standard QKD by placing the state source near the untrusted device.

The Gist

Imagine you and a friend want to share a secret code (a "key") to lock your messages so no one else can read them. In the world of quantum physics, we have a special way to do this called Quantum Key Distribution (QKD). It's like a magic lock that breaks if anyone tries to peek at it, thanks to the laws of physics.

However, there's a catch: to make this work, you usually have to trust that your lock and your friend's lock are perfect. If the locks are slightly broken or made by a shady manufacturer, a hacker could sneak in and steal your secret without you knowing.

The Problem: The "Perfect Lock" Dilemma

Scientists have tried to solve this by creating Device-Independent (DI) protocols. Think of this as a system where you don't need to trust the locks at all; you just check if the magic is working. But there's a big downside: these "trust-nothing" systems are incredibly fragile. They need to catch almost every single particle of light (photon) they send. If even a few get lost in the fiber-optic cables (which happens over long distances), the system fails. It's like trying to play a game of catch in a hurricane; if the ball gets lost too often, you can't play.

The Solution: The "Semi-Trusted" Compromise

This paper introduces a clever middle ground called One-Sided Device-Independent (1SDI) QKD.

Imagine a scenario where you (Alice) have a high-tech, certified, and trusted lock in your secure lab. Your friend (Bob), however, is using a "black box" lock that might be old, broken, or even built by a potential hacker.

• The Old Way: In standard setups, the "black box" side had to be perfect, or the whole thing failed.
• The New Way: The authors show that as long as your side is trusted, Bob's side can be a bit messy. Even if Bob's detector is missing more than half the balls (light particles), you can still generate a secret key safely.

The Big Breakthrough: The 50% Magic Line

The most exciting part of this paper is the number 50.1%.

Think of Bob's detector as a net trying to catch fish.

• If the net has holes so big that it catches less than 50% of the fish, it's impossible to prove the fish are real (secure).
• The authors proved that if Bob's net catches just over 50% of the fish, you can still generate a secret key that is mathematically guaranteed to be safe, even if the hacker is using the most sophisticated tricks (called "coherent attacks").

This is a huge deal because 50% is the theoretical limit. You can't really go much lower than that. They managed to hit almost the absolute bottom limit of what is physically possible.

How They Did It: The "No-Filter" Rule

Previous attempts to make this work used a trick called "post-selection." This is like saying, "We will only count the rounds where Bob caught a fish, and we'll throw away all the rounds where he missed."

• The Flaw: If you throw away data, a clever hacker can hide their cheating in the "missed" rounds.
• The Fix: This paper says, "Don't throw anything away!" Even if Bob's detector clicks nothing (a "miss"), you keep that data in the record. By analyzing the entire picture—including the misses—they proved the system is secure against the most general types of attacks.

The Distance Test: How Far Can We Go?

The authors also asked: "How far can we send this secret key?"
In standard "trust-nothing" systems, the distance is very short because light gets lost in the cables. But in this new setup, they placed the "source of the light" right next to Bob's lab.

• The Setup: The light starts near Bob (so he doesn't lose any), and travels a long way to Alice.
• The Result: They calculated that with current technology, this system could securely send keys over 247 kilometers (about 153 miles). This is comparable to the distances used in standard, less secure systems.

The Real-World Picture

Imagine a bank (Alice) and a remote server (Bob).

• The bank trusts its own equipment completely.
• The server is in a different city, and its equipment might be old or maintained by a third party.
• Using this new method, the bank can still establish a super-secure connection with the server, even if the server's equipment is only 50% efficient at catching signals.

Summary

This paper takes an old, famous protocol (BB84) and updates it with modern math to prove it works even when one side is unreliable.

1. Trust only one side: You need to trust your own device, but you don't need to trust the other person's.
2. Low efficiency is okay: The other person's device only needs to work 50.1% of the time.
3. No data dumping: You keep all the data, even the "failures," which makes the system safe against smart hackers.
4. Long distances: It can work over hundreds of kilometers, making it practical for real-world use.

In short, they found a way to make quantum security robust enough to survive the real world, where equipment isn't perfect and distances are long.

Technical Summary

Technical Summary: One-sided DI-QKD Secure Against Coherent Attacks Over Long Distances

Problem Statement
Quantum Key Distribution (QKD) offers provably secure communication but relies on strict assumptions regarding device behavior, creating potential security loopholes if devices deviate from their mathematical models. Device-Independent (DI) QKD protocols address this by minimizing device assumptions; however, they are currently limited in transmission distance due to extreme sensitivity to photon losses, requiring detection efficiencies that are difficult to achieve over long fibers. Conversely, standard Device-Dependent QKD is robust over distance but lacks security against device imperfections. There is a need for protocols that balance realistic device assumptions with long-distance implementability while maintaining security against the most general class of attacks, specifically coherent attacks.

Methodology
The authors propose a One-Sided Device-Independent (1SDI) QKD protocol based on an entanglement-based version of the BB84 scheme. The scenario involves two parties, Alice and Bob, sharing an entangled state.

• Trust Model: Alice's device is "semi-trusted" (trusted to perform two anti-commuting observables with discrete outcomes, though the Hilbert space dimension is unrestricted), while Bob's device and the state source are completely untrusted.
• Configuration: To mitigate losses, the untrusted source is positioned near Bob's laboratory. This minimizes transmission loss to the untrusted side, while losses on Alice's trusted side are handled by discarding non-detection rounds (fair-sampling assumption).
• Security Framework: Unlike previous 1SDI protocols that rely on post-selection of outcomes from the untrusted party (which invalidates security proofs against coherent attacks), this protocol retains all of Bob's outcomes, including non-detection events ($\emptyset$). This allows the application of the Generalized Entropy Accumulation Theorem (GEAT) to prove security against coherent attacks.
• Key Rate Analysis: The asymptotic key rate is lower-bounded using the Devetak-Winter formula: $r_\infty \ge H(A_1|E) - H(A_1|B_1)$.
  • $H(A_1|B_1)$ is computed from observed statistics.
  • $H(A_1|E)$ (the conditional Von Neumann entropy bounding Eve's information) is estimated using two techniques:
    1. Analytical: Based on bounds derived for two-dimensional spaces and extended to arbitrary dimensions via Jordan's lemma.
    2. Numerical: Utilizing the Brown-Fawzi-Fawzi (BFF) technique, a semi-definite programming (SDP) approach based on the NPA hierarchy, which optimizes the entropy bound using the full probability distribution $p(a,b|x,y)$.

Key Contributions

1. Coherent Attack Security: The work demonstrates that a 1SDI protocol with two measurements per party is secure against coherent attacks without performing post-selection on the untrusted side. This is achieved by satisfying the non-signaling constraints required by the GEAT.
2. Detection Efficiency Threshold: The authors establish a security threshold for Bob's detection efficiency on the untrusted side of 50.1%. This is nearly the theoretical limit for protocols with two untrusted measurements, as any protocol with efficiency below 50% allows for joint measurability of Bob's observables, rendering the protocol insecure.
3. Long-Distance Feasibility: By placing the source near the untrusted party (Bob), the protocol minimizes losses on the critical untrusted channel. The authors estimate that, under realistic experimental parameters (90% total detection efficiency for Bob, 99% state visibility, and dark count rate of $10^{-6}$), the protocol can be implemented over distances up to 247 km.
4. Optimization Strategies: The study explores the use of partially entangled states and noisy pre-processing. While maximally entangled states were found to be optimal for the specific 50.1% threshold case, the analysis shows that for lower visibilities, partially entangled states combined with noisy pre-processing can improve the key rate.

Results

• Key Rate vs. Efficiency: Using the BFF numerical technique with full statistics (keeping Bob's non-detection as a separate outcome), the protocol maintains a positive key rate up to a detection efficiency of 50.1%.
• Noise Robustness: The protocol's robustness against depolarizing noise and dark counts was analyzed. The results indicate that while the joint measurability of Bob's observables sets a hard boundary for insecurity, there exists a region where the key rate is non-positive even when measurements are not jointly measurable, suggesting room for further protocol refinement.
• Distance Estimation: Under the assumption that the source is co-located with Bob, the protocol achieves a positive key rate at approximately 247 km, a distance comparable to standard QKD protocols, provided the untrusted side maintains high detection efficiency.

Significance
The paper claims to bridge the gap between the high security of Device-Independent protocols and the long-distance capability of standard QKD. By applying recent theoretical frameworks (GEAT and BFF techniques) to a modified BB84 scenario, the authors demonstrate that 1SDI QKD can be made secure against coherent attacks with detection efficiencies that are within current experimental capabilities. The work suggests that the theoretical limit for two-measurement protocols (50%) is practically achievable, offering a viable path for long-distance, high-security quantum communication where one party's device can be trusted to a specific degree (semi-trusted) while the other remains untrusted. The authors note that these results have been independently confirmed in subsequent works utilizing local Bell tests.