USD-Based Intercept-Resend in QKD Protocols

• The paper investigates USD-based intercept-resend strategies that use unambiguous state discrimination on non-orthogonal BB84 signals to enable error-free conclusive events while altering gain and yield statistics.
• It details how the adversary replaces intercepted pulses with accurately prepared states on conclusive outcomes and blocks or dummy-resends on inconclusive results, thereby shifting disturbance from QBER to yield patterns.
• Implementation through laser damage and beam splitter modifications exposes vulnerabilities in decoy-state QKD, emphasizing the need for rigorous countermeasures and protocol integrity checks.

Searching arXiv for the cited papers and closely related work on intercept–resend, USD, and decoy-state QKD. A USD-based intercept-resend strategy is an intercept–measure–reprepare attack in which the adversary replaces minimum-error guessing by a measurement with conclusive and inconclusive outcomes: on a conclusive result the signal state is identified without error and a corresponding state is resent, while on an inconclusive result the pulse is blocked or replaced by a dummy state. In BB84-derived systems this attack targets non-orthogonal signal ensembles, so it exchanges the disturbance typical of naive intercept-resend for a different signature: zero error on the forwarded subset, but modified yields, gains, or loss statistics. In the recent arXiv literature, the attack appears both explicitly, in damaged decoy-state BB84, and implicitly, as the natural USD analogue of state-discrimination-based intercept-resend models used in quantum secure imaging and related protocols (Zhao et al., 2024, Sushchev et al., 21 Jul 2025).

1. Definition within the intercept–resend family

In the standard intercept–resend paradigm, Eve intercepts a quantum signal, performs a measurement, infers a classical description of the state, and resends a replacement state to Bob. A USD-based version specializes the measurement stage. Instead of using a minimum-error measurement, Eve attempts unambiguous state discrimination on the signal ensemble. The defining feature is that conclusive outcomes are error-free, whereas inconclusive outcomes occur with nonzero probability.

For phase-encoded BB84, the relevant states are non-orthogonal. In the quantum secure imaging construction based on phase-encoding and weak+vacuum decoy-state BB84, the $n$-photon states $\{|n_k\rangle\}_{k=0}^3$ are symmetric phase-encoded states, and the paper makes the conceptual point that any measurement trying to identify $k$ must either make errors or avoid errors at the cost of inconclusive outcomes. It further states that the analyzed Curty–Lütkenhaus “filter + SRM” attack is minimum-error discrimination with filtering, not USD per se, but that both are governed by the same overlaps and therefore belong to the same state-discrimination-based attack family (Zhao et al., 2024).

A canonical USD-based intercept-resend therefore has four stages. Eve first performs a photon-number-resolving step, or in the idealized model a QND measurement to obtain the total photon number $n$. She then attempts USD on the set $\{|n_k\rangle\}_{k=0}^3$. If the outcome is conclusive, she prepares a new pulse in the corresponding BB84 state and sends it to Bob. If the outcome is inconclusive, she either blocks the pulse by sending vacuum or replaces it with some dummy state. The critical consequence is structural rather than merely probabilistic: disturbance can be shifted away from QBER on conclusive events and into the observed gain pattern.

USD is possible only when the candidate states are linearly independent. The quantum secure imaging paper states this condition explicitly and notes that the maximum conclusive probability depends on the geometric properties of the state set through the Gram matrix, with $P_{\text{USD}}<1$ whenever the relevant pairs are nonorthogonal (Zhao et al., 2024).

2. Signal ensembles and protocol mechanics

The most detailed explicit USD construction in the supplied corpus occurs in decoy-state BB84 after laser damage attack. There, the underlying source is a phase-randomized weak coherent source with signal, decoy, and vacuum intensities. Security relies on fixed and stable mean photon numbers, but the paper models laser damage as a multiplicative alteration

$$\tilde{\mu}=\kappa\mu,\qquad \tilde{\nu}=\kappa\nu,$$

so that the actual intensities can be pushed outside the faint-pulse regime (Sushchev et al., 21 Jul 2025).

In that setting Eve does not try to discriminate intensities. She performs a USD-type measurement on the BB84 state itself by using a Bob-like passive-basis receiver with threshold detectors. A conclusive outcome is defined as the event in which exactly three detectors click. The paper derives the yield for an $n$-photon Fock component,

$Y_n^{\rm USD}=1-\frac{3^n-1}{2^{2n-1}},$

and, for coherent states with LDA-modified mean photon numbers, the corresponding gains

$$Q_\mu=\left(1-e^{-\tilde{\mu}/2}\right)\left(1-e^{-\tilde{\mu}/4}\right)^2,\qquad Q_\nu=\left(1-e^{-\tilde{\nu}/2}\right)\left(1-e^{-\tilde{\nu}/4}\right)^2.$$

On a conclusive event, Eve resends a bright pulse or another suitably chosen state; on an inconclusive event, she sends vacuum. The resulting channel is therefore not an honest lossy channel but one whose effective yields are entirely set by Eve’s conclusive probabilities (Sushchev et al., 21 Jul 2025).

The same logic is present in the quantum secure imaging paper, though there USD is discussed as a conceptual extension rather than the directly analyzed attack. The transmitted phase-encoded weak coherent pulses are used simultaneously for BB84 key generation and for computational ghost imaging via a DMD-defined intensity pattern $\{|n_k\rangle\}_{k=0}^3$0. In this architecture, a jammer at the target can intercept the phase-encoded weak coherent pulses, measure them, and resend manipulated signals toward Bob, so a USD-based intercept-resend is simultaneously a key-extraction strategy and an image-spoofing strategy (Zhao et al., 2024).

The modified USD construction of the decoy-state vulnerability paper adds a beam splitter of transmittance $\{|n_k\rangle\}_{k=0}^3$1 before Eve’s USD apparatus. This suppresses higher-photon-number components and changes the yields to

$\{|n_k\rangle\}_{k=0}^3$2

The purpose is not merely implementation convenience: it reshapes the photon-number dependence so that the induced decoy statistics more closely resemble a three-photon-dominated process, lowering the required attenuation damage threshold at the price of lower overall gain (Sushchev et al., 21 Jul 2025).

3. Information–disturbance trade-offs and statistical signatures

The central theoretical distinction between USD-based and minimum-error intercept-resend is the location of the unavoidable disturbance. Minimum-error strategies introduce errors directly; USD can suppress error on the forwarded subset, but only by paying in inconclusive events and therefore in altered transmission statistics. The quantum secure imaging analysis states this directly: zero error on conclusive results is accompanied by nonunit conclusive probability, so many pulses must be discarded or modified, which impacts the detection rates (Zhao et al., 2024).

The same paper quantifies the disturbance floor for the related Curty–Lütkenhaus filter+SRM attack. For phase-encoded BB84 multiphoton components it derives

$\{|n_k\rangle\}_{k=0}^3$3

and combines these with decoy-state inequalities to obtain a lower bound on the decoy QBER under intercept-resend,

$\{|n_k\rangle\}_{k=0}^3$4

Using the experimental parameters $\{|n_k\rangle\}_{k=0}^3$5, $\{|n_k\rangle\}_{k=0}^3$6, $\{|n_k\rangle\}_{k=0}^3$7, $\{|n_k\rangle\}_{k=0}^3$8, and $\{|n_k\rangle\}_{k=0}^3$9, the measured decoy QBER was $k$0, far below $k$1, and the secure key rate during imaging was $k$2 bps. Within that framework, any attack in the same state-discrimination class, including a USD variant, must either respect the disturbance bound and become detectable or produce inconsistent gain statistics (Zhao et al., 2024).

A related information-theoretic perspective appears in the depolarizing-channel analysis of sequential intercept-resend. There the security condition is expressed as

$k$3

with $k$4 the binary Shannon entropy of the channel error. That paper does not implement USD, but it explicitly remarks that the same framework applies to a USD-based strategy once the corresponding conditional probabilities $k$5 and $k$6 are supplied. The relevant point is that background noise can mask or redistribute intercept-resend disturbance, so the operative quantity is not raw QBER alone but the joint relation among QBER, mutual information, and channel statistics (Dehmani et al., 2013).

The trade-off is therefore protocol dependent. In an idealized decoy-state analysis, USD tends to reveal itself through anomalous gains. In a noisy channel it can hide behind pre-existing disturbance more effectively, but then the security test must move from a single error threshold to a joint comparison of Bob’s and Eve’s information with the entropy contribution of the channel.

4. Realistic decoy-state break enabled by laser damage

The most concrete modern formulation of a USD-based intercept-resend strategy is the combination of laser damage attack and USD against decoy-state BB84. The attack begins by using high-power illumination to alter Alice’s attenuation so that the actual intensities $k$7 exceed the regime assumed by the decoy-state proof. Once this alteration passes a critical threshold on the order of $k$8 dB, the three-click USD measurement becomes efficient enough that Eve can obtain the entire secret key while still reproducing decoy statistics that pass standard checks (Sushchev et al., 21 Jul 2025).

The paper defines the critical alteration $k$9 as the point where the decoy lower bound on the single-photon yield first becomes positive. Its approximate analytic threshold is

$n$0

For representative decoy parameters it reports the following numerical USD thresholds: $n$1 dB for $n$2; $n$3 dB for $n$4; and $n$5 dB for $n$6. These values fall within or near experimentally demonstrated attenuation reductions for realistic components (Sushchev et al., 21 Jul 2025).

The modified USD setup with an extra beam splitter lowers the threshold further. Its analytic estimate is

$n$7

which approaches the three-photon PNS threshold as $n$8. The countervailing effect is a strong gain penalty; in the small-$n$9 limit,

$\{|n_k\rangle\}_{k=0}^3$0

The paper therefore presents the attack as an implementation-specific break, not a contradiction of decoy-state theory under its own assumptions. Stable attenuation is a hidden hardware assumption; once it is violated, the attack statistics can be made consistent with a positive secret key rate even though Eve holds the full raw key (Sushchev et al., 21 Jul 2025).

This case is especially important because it demonstrates that a USD-based intercept-resend strategy can be technologically realistic without QND measurements or quantum memory. The enabling resource is not a fundamentally stronger discrimination theorem but a change in the operating point of the source.

5. Detection mechanisms and countermeasures across protocols

Different protocols expose different observables to a USD-based intercept-resend strategy. The following contexts illustrate the range of signatures and defenses.

Context	Role of USD/intercept-resend	Observable consequence
Decoy-state QSI	Conclusive forwarding, inconclusive blocking or replacement	QBER/gain inconsistency; $\{|n_k\rangle\}_{k=0}^3$1 bound
LDA-damaged decoy-state BB84	Three-click USD with threshold detectors	Standard decoy checks can be misled
Bidirectional QKD with randomized ancilla	Any faked-state resend, regardless of Eve’s measurement	Alert detectors click with unavoidable probability
Mach–Zehnder surveillance	Any which-path discrimination, including USD-like resend	Bright-port flux reduced to at most $\{|n_k\rangle\}_{k=0}^3$2

In bidirectional phase-encoded QKD with a randomized ancillary polarization qubit, the defense is architectural rather than statistical. Bob applies a random SU(2) polarization transformation $\{|n_k\rangle\}_{k=0}^3$3 to outgoing photons. Genuine photons traverse the full Bob $\{|n_k\rangle\}_{k=0}^3$4 Alice $\{|n_k\rangle\}_{k=0}^3$5 Bob loop and the ancillary polarization is effectively undone by the Faraday mirror and second pass, but Eve’s injected faked state sees the randomization only once. The paper states that this makes it impossible for Eve to avoid triggering the alert, no matter what faked-state of light she uses. Since the attack model is explicitly framed to include general intercept-resend with faked-state photons, a USD front-end changes only Eve’s measurement fidelity; it does not change the fact that her resent light is routed into alert detectors with substantial and unavoidable probability (Hegazy et al., 2022).

In Mach–Zehnder-based quantum-secured surveillance, any intercept-resend attempt that determines whether a photon is present in the fence branch destroys the coherence needed for interference. The paper states that a single photon arriving at the second beam splitter from only one arm has only a $\{|n_k\rangle\}_{k=0}^3$6 chance of arriving at the correct detector, leading to an average output flux at the bright port equal to $\{|n_k\rangle\}_{k=0}^3$7. A USD-based intercept-resend is a special case of this general measurement–resend logic: it may produce conclusive outcomes on some events, but the corresponding which-path information reduces the bright-port flux and therefore remains detectable through interference loss (Bishop et al., 2013).

Counterfactual quantum cryptography yields yet another pattern. There the general intercept-resend analysis is broad enough to include any POVM on the traveling mode, so a USD-based strategy is already subsumed by the general unitary $\{|n_k\rangle\}_{k=0}^3$8 model. The key rate is bounded by observed probabilities $\{|n_k\rangle\}_{k=0}^3$9 and error rates $P_{\text{USD}}<1$0; under ideal devices, a USD-based attack cannot give Eve more information than the general bound allows. Practical imperfections such as time-shift and detector-efficiency mismatch can, however, amplify the effectiveness of intercept-resend strategies by corrupting the observable statistics on which the proof relies (Zhang et al., 2011).

The imaging-specific case closes the circle. In quantum secure imaging, spoofing the image requires coherent manipulation of the pattern-dependent counts $P_{\text{USD}}<1$1, not merely opportunistic interception of isolated pulses. The paper’s claim is that such tampering necessarily pushes either QBER or decoy-state gains outside the safe region, so authenticity of imaging information is enforced through the same observables that certify key security (Zhao et al., 2024).

6. Generalizations, limitations, and broader protocol dependence

Not every paper in the corpus analyzes USD explicitly, but several provide formalisms into which USD-based intercept-resend can be inserted. Under collective rotation noise, for example, the BB84 analysis is written in terms of conditional probability matrices $P_{\text{USD}}<1$2 and $P_{\text{USD}}<1$3. The paper states that this framework can incorporate any intercept-resend strategy, including USD, by replacing Eve’s measurement model and resend map. It also reports that, for projective intercept-resend, Eve’s mutual information is approximately $P_{\text{USD}}<1$4 bits at $P_{\text{USD}}<1$5 and $P_{\text{USD}}<1$6, with a minimum around $P_{\text{USD}}<1$7 bits at $P_{\text{USD}}<1$8. This suggests that, for some noise models, the operative question is not whether USD exists but how the channel deformation changes the distinguishability of the rotated state ensemble (Masood et al., 20 May 2026).

High-dimensional or continuously parameterized signal families further constrain USD. In DTQW-based quantum direct communication, the traveling states are discrete-time quantum walk states with position-space superposition and coin–position entanglement. The paper argues that Eve cannot determine the incoming state by measurement alone and reports that, for odd $P_{\text{USD}}<1$9, suitable $$\tilde{\mu}=\kappa\mu,\qquad \tilde{\nu}=\kappa\nu,$$0, and large $$\tilde{\mu}=\kappa\mu,\qquad \tilde{\nu}=\kappa\nu,$$1, the mutual information under intercept-resend drops well below the $$\tilde{\mu}=\kappa\mu,\qquad \tilde{\nu}=\kappa\nu,$$2 value of LM05/DL04, in some cases below $$\tilde{\mu}=\kappa\mu,\qquad \tilde{\nu}=\kappa\nu,$$3. This suggests that large, structured, randomly parameterized state alphabets can make any state-discrimination-based intercept-resend, including a USD variant, much less effective unless Eve can also invert hidden dynamics (S et al., 2020).

The multi-photon double-lock protocol makes this point explicitly in USD language. It observes that practical implementations may restrict the polarization rotations to a finite set and that such limitation may open up the unambiguous state discrimination attack. Its stated mitigation is to use an extremely large set of polarization rotations, changed frequently and kept local to Alice and Bob, so that the number of photons needed must be greater than or equal to the number of polarization states in the middle of the three-stage protocol (Chan et al., 2015).

By contrast, some protocols remain comparatively weak even before USD is introduced. The analysis of Bub’s pre- and post-selection protocol treats only projective intercept-resend and translucent attacks and concludes that the protocol is weaker than BB84 against both. Since the backward-leg signal alphabet is the BB84 four-state set, full USD on that set is not available in a two-dimensional Hilbert space, so the main lesson there is not that USD adds a new failure mode, but that projective intercept-resend already yields an unfavorable information–disturbance trade-off (Azuma et al., 2018).

Taken together, these results define a precise scope for the term. A USD-based intercept-resend strategy is not a single attack but a family of state-discrimination attacks whose effectiveness depends on three layers of structure: the geometry of the signal ensemble, the auxiliary observables monitored by the protocol, and the integrity of the implementation assumptions. In ideal decoy-state analyses it is usually exposed by altered yields; in imaging and interferometric systems it is exposed by QBER, gains, or coherence loss; in some hardware-compromised decoy-state systems it can become fully practical and pass standard checks; and in architectures with randomized ancillary degrees of freedom it can be neutralized even when Eve’s measurement stage is arbitrarily strong.