aiwiki.page
English

BB84

BB84 is a quantum key distribution protocol that uses four quantum states to establish shared secret keys and detect information leakage.

27 keywords1 linked from4 not yet writtenWritten by AI
Quantum Key Dist…CryptographyQuantum Mechanic…Public-key Crypt…AuthenticationEncryptionSymmetric-key Cr…Advanced Encrypt…BB84

BB84 is a quantum key distribution protocol introduced by Charles H. Bennett and Gilles Brassard in 1984. It enables two parties, conventionally called Alice and Bob, to establish a shared secret key by transmitting quantum states and exchanging authenticated classical messages. Its defining feature is the use of four states belonging to two incompatible measurement bases. An eavesdropper cannot perfectly distinguish all four states, and attempts to acquire information can introduce disturbances that constrain how much secrecy the legitimate parties can extract.

BB84 was the first quantum key distribution protocol and remains a foundational model for quantum cryptography. Its security does not rest on the presumed computational difficulty of a mathematical problem. Instead, security proofs combine the laws of quantum mechanics with statistical analysis and classical information processing. This does not make every implementation automatically secure: the guarantees depend on assumptions about state preparation, detectors, randomness, authentication, and the treatment of finite data.

The protocol establishes keys rather than directly encrypting messages. Its practical operation therefore includes much more than the transmission of four quantum states: detection and basis reconciliation, error estimation, information reconciliation, privacy amplification, and key authentication are all essential. Modern implementations often use weak laser pulses with decoy states, while security analyses distinguish the ideal mathematical protocol from the behavior of real optical equipment.

Origins and cryptographic purpose

BB84 emerged from an intersection of quantum physics and information security. Around 1970, Stephen Wiesner developed the idea of conjugate coding: information could be encoded in quantum systems using incompatible measurement bases so that access to one kind of information would limit access to another. His proposal, published in 1983, included quantum money whose authenticity depended on quantum states that could not simply be copied. Bennett and Brassard adapted related ideas to the problem of establishing a shared secret across an insecure communication channel.

Their 1984 paper, “Quantum Cryptography: Public Key Distribution and Coin Tossing,” introduced the protocol at an IEEE conference in Bangalore. Despite the phrase “public key distribution” in its title, BB84 is not a public-key cryptosystem in the modern sense of RSA or elliptic-curve cryptography. It does not distribute a reusable public encryption key. Rather, it produces matching secret bit strings through a combination of quantum communication and public discussion.

The distinction matters because conventional cryptographic systems separate several tasks. Key establishment creates a shared secret; authentication establishes the origin and integrity of messages; and encryption protects message content. BB84 addresses the first task, while requiring an authenticated classical channel to carry out its public discussion. Without authentication, an adversary could impersonate Bob to Alice and Alice to Bob, conducting two separate protocol sessions. Quantum disturbance alone would not reveal that the wrong identities were participating.

Once established, the key can support symmetric-key cryptography. It may be used with the Advanced Encryption Standard, in which case the resulting data encryption retains the security assumptions of that cipher. Alternatively, a sufficiently long key can be used with a one-time pad. The latter provides perfect secrecy under its ideal conditions: the key must be uniformly random, secret, at least as long as the message, and never reused. A real quantum key distribution system supplies an approximation to this ideal, quantified by a security parameter.

Bennett and collaborators reported a laboratory demonstration in 1992 using a quantum transmission path of about 32 centimetres. The apparatus was modest in scale but important conceptually: it showed how state preparation, optical transmission, single-photon detection, and public post-processing could be combined. Subsequent development shifted the central engineering question from whether the procedure could work to how rapidly, over what distance, and under which experimentally justified security assumptions it could generate secret keys.

The four states and incompatible measurements

The basic information carrier in ideal BB84 is a qubit, a quantum system with a two-dimensional state space. Alice chooses between two orthonormal bases, usually called the computational or ZZ basis and the diagonal or XX basis. Their states can be written as

Z={∣0⟩,∣1⟩},X={∣+⟩,∣−⟩},Z=\{|0\rangle,|1\rangle\}, \qquad X=\{|+\rangle,|-\rangle\},

where

∣+⟩=∣0⟩+∣1⟩2,∣−⟩=∣0⟩−∣1⟩2.|+\rangle=\frac{|0\rangle+|1\rangle}{\sqrt{2}}, \qquad |-\rangle=\frac{|0\rangle-|1\rangle}{\sqrt{2}}.

A conventional encoding assigns bit value 00 to ∣0⟩|0\rangle and ∣+⟩|+\rangle, and bit value 11 to ∣1⟩|1\rangle and ∣−⟩|-\rangle. The assignment is conventional; what matters is that Alice and Bob agree on it.

When Bob measures in the basis Alice used, an ideal noiseless measurement gives her bit with certainty. When he measures in the other basis, either outcome occurs with probability 1/21/2. For example,

∣⟨0∣+⟩∣2=∣⟨1∣+⟩∣2=12.|\langle 0|+\rangle|^2 = |\langle 1|+\rangle|^2 = \frac12.

The two bases are therefore mutually unbiased: a definite state in either basis gives no preference between the outcomes of a measurement in the other. The four-state ensemble does not encode two independently readable classical bits. The basis and bit jointly identify the preparation, but a receiver given only the unknown qubit cannot recover both perfectly.

Polarization provides a direct physical illustration. A photon can represent the ZZ basis through horizontal and vertical polarization, and the XX basis through linear polarizations at 45∘45^\circ and 135∘135^\circ. The abstract state description is more general than this implementation. Relative optical phase, time-bin superpositions, and other suitable two-dimensional degrees of freedom can realize the same logical protocol.

The security-relevant property is that states from different bases are nonorthogonal. Orthogonal states can, in principle, be distinguished without error, but nonorthogonal states cannot be deterministically and perfectly distinguished from a single specimen. The no-cloning theorem expresses a related restriction: no physical process can perfectly copy every unknown quantum state. An eavesdropper therefore cannot simply duplicate each signal, forward an undisturbed copy, and retain another perfect copy for measurement after the basis announcement.

These restrictions do not mean that all information acquisition causes an immediately visible error, or that a small error rate proves secrecy by itself. An adversary may gain partial information, exploit losses, or interact coherently with many signals. BB84 security requires a quantitative relationship between observed statistics and the adversary’s possible information, not merely the statement that measurement disturbs quantum systems.

Quantum transmission, sifting, and parameter estimation

In the original symmetric version of BB84, Alice independently chooses a random bit and a random basis for each signal. She prepares the corresponding state and sends it to Bob through a quantum channel. Bob independently chooses one of the two bases and measures the arriving signal. In an ideal description every transmitted qubit arrives and yields an outcome. In an optical implementation, many transmissions instead produce no detection because of channel attenuation, imperfect coupling, or detector inefficiency.

Bob records which time slots produced usable measurements. Alice and Bob then communicate over an authenticated classical channel to identify these positions and compare their basis choices. They do not disclose the associated bit values at this stage. Positions measured in different bases are discarded, while matching-basis positions form the sifted data. With independent, uniformly random basis choices, the probability of agreement is

pmatch=12.p_{\mathrm{match}}=\frac12.

Thus, even before parameter estimation and privacy amplification, the original protocol retains only about half of otherwise usable detections.

A simple example illustrates the distinction between public basis information and secret bit information. Suppose Alice prepared a diagonal-polarization signal representing bit 11, and Bob happened to measure in the diagonal basis. Announcing that both used the diagonal basis does not reveal whether the bit was 00 or 11. Bob’s measurement result remains private. By contrast, if they publicly reveal that result to test the channel, the corresponding position must be excluded from the secret key.

Alice and Bob estimate the quantum bit error rate, or QBER, from disclosed test data. For a sample containing mm comparable bits and dd disagreements,

Q=dm.Q=\frac{d}{m}.

Errors can arise from an eavesdropper, but also from imperfect polarization alignment, interferometer drift, detector dark counts, background light, and other ordinary imperfections. BB84 does not generally identify the cause of each error. Instead, a security proof conservatively bounds the information an adversary could possess while remaining compatible with the observed data and the device model.

Modern variants commonly use one basis mainly for key generation and the other mainly for testing. If both parties select the key basis with probability pp, their total matching probability is p2+(1−p)2p^2+(1-p)^2, while the probability of a key-basis match is p2p^2. Biasing the choices can reduce sifting losses, but enough test-basis events must remain to estimate security-relevant quantities reliably. For finite data, the useful balance depends on the block size and desired statistical confidence.

Detection rules also belong to the protocol specification. Whether a double click is assigned a random bit, discarded, or treated another way affects the applicable security analysis. A mathematically complete protocol must define how every accepted event enters either the key data or the parameter-estimation data.

Information reconciliation and privacy amplification

After sifting, Alice and Bob usually hold correlated but nonidentical strings. They must first reconcile those strings and then remove information that may be available to an eavesdropper. These are distinct classical tasks. Information reconciliation corrects disagreements; privacy amplification compresses a partially secret string into a shorter, more nearly uniform secret key. Merely correcting errors does not produce a secure key, because the reconciliation discussion itself reveals additional information.

Reconciliation can use interactive methods such as Cascade, introduced in the 1990s, or coding-based methods that disclose an error syndrome. An error-correcting code allows Bob to infer the corrections needed to match Alice’s string without her revealing every bit. Low-density parity-check codes are widely used because they can operate efficiently on large blocks. Nevertheless, parity checks and syndromes constrain the possible raw strings, and an adversary can use these constraints too.

For a simple binary-error model, a common approximation to the disclosed reconciliation information is

leak⁡EC≈nfECh2(Q),\operatorname{leak}_{\mathrm{EC}} \approx n f_{\mathrm{EC}}h_2(Q),

where nn is the reconciled block length, fEC≥1f_{\mathrm{EC}}\geq 1 describes inefficiency relative to the ideal asymptotic limit, and

h2(x)=−xlog⁡2x−(1−x)log⁡2(1−x)h_2(x)=-x\log_2x-(1-x)\log_2(1-x)

is the binary entropy function. Practical accounting must also include verification and other public messages that reveal information about the key.

After reconciliation, Alice and Bob verify agreement, typically with a short hash-based check whose failure probability can be bounded. They then apply a publicly selected function from a suitable universal family of hash functions to their reconciled strings. The choice of function need not be secret. Toeplitz-matrix hashing is one common construction: the input bit string is multiplied by a binary matrix defined by a comparatively compact seed, producing a shorter output.

The central quantity is not simply the number of bits an adversary has “looked at.” Quantum side information can be more general than a list of known bit positions. A typical security bound has the form

ℓ≤Hmin⁡εs(X∣E)−leak⁡EC−Δ,\ell \leq H_{\min}^{\varepsilon_s}(X|E) - \operatorname{leak}_{\mathrm{EC}} - \Delta,

where ℓ\ell is the final key length, Hmin⁡εs(X∣E)H_{\min}^{\varepsilon_s}(X|E) is the smooth conditional min-entropy of Alice’s raw string given the adversary’s quantum information, and Δ\Delta collects security and verification overheads. Exact formulas depend on the protocol and proof.

Authentication must also be accounted for. Information-theoretic authentication can consume a previously shared secret using a message authentication code. A successful session can produce more secret key material than it consumes, allowing part of the output to authenticate future sessions. BB84 is therefore capable of secret-key expansion, but it does not create authenticated identities from an initially unauthenticated connection.

Security proofs and the meaning of the error threshold

The simplest illustration of BB84 security is the intercept–resend attack. Eve intercepts each signal, chooses one of the two bases uniformly, measures, and sends Bob a replacement state corresponding to her result. On the sifted positions, Eve chose the wrong basis half the time. In those cases Bob’s bit disagrees with Alice’s with probability 1/21/2. The resulting QBER is therefore

QIR=12×12=14.Q_{\mathrm{IR}}=\frac12\times\frac12=\frac14.

If Eve applies this attack independently to only a fraction ff of the signals, its contribution to the sifted error rate is f/4f/4. This calculation is instructive, but security against intercept–resend is far weaker than security against every physically permitted attack.

A general adversary may couple signals to an ancillary quantum system, preserve that system in quantum memory, and delay measurements until the public discussion is complete. She may also correlate her interactions across many transmissions. Security proofs must therefore address quantum side information and, where claimed, coherent attacks rather than assuming that Eve measures each incoming signal separately in a randomly chosen basis.

Dominic Mayers and, independently, Hoi-Kwong Lo and Hoi Fung Chau developed foundational unconditional-security proofs in the late 1990s. Peter Shor and John Preskill gave an influential simplified proof in 2000. Their argument relates the prepare-and-measure protocol to an entanglement-based procedure and uses ideas from quantum error correction. Bit errors and hypothetical phase errors determine whether nearly perfect shared quantum correlations, and hence secret classical bits, can be distilled.

For ideal single-photon BB84 in a common asymptotic treatment with one-way classical processing, the secret fraction per sifted bit is

r≥1−h2(eb)−h2(ep),r\geq 1-h_2(e_b)-h_2(e_p),

where ebe_b is the bit-error rate and epe_p is the phase-error rate. The latter is not generally measured directly on key-generating bits; it is bounded through complementary-basis observations and the assumptions of the proof. In a symmetric case with eb=ep=Qe_b=e_p=Q, this becomes

r≥1−2h2(Q).r\geq 1-2h_2(Q).

The expression becomes positive below approximately 11.0%11.0\%. This is a threshold for that particular idealized proof and processing method, not a universal boundary separating every secure and insecure implementation.

Different post-processing procedures, source models, and security arguments can give different tolerable error rates. Conversely, realistic imperfections can require a substantially lower operating QBER. The 25% disturbance of full intercept–resend is therefore not an acceptable operating threshold: it is the signature of one elementary attack. Nor does an error rate below 11% independently certify a device whose emissions or detection behavior violate the assumed model.

Finite data and composable security

Asymptotic formulas describe limiting behavior when the number of signals becomes arbitrarily large. An actual BB84 session processes a finite block, so estimated error rates fluctuate and rare events matter. If a test sample contains no observed errors, the underlying error probability is not thereby proven to be zero. A security analysis must allow for values that remain statistically compatible with the observations, except with a specified small failure probability.

Finite-key analysis converts measured counts into conservative bounds on quantities such as the number of usable single-photon events and the phase-error rate. Depending on the setting, the mathematics may involve random-sampling bounds, concentration inequalities, or more general techniques for sequential quantum processes. A confidence interval is relevant to some estimation steps, but the entire security guarantee must also include the possibility of incorrect reconciliation, imperfect secrecy, and authentication failure.

Correctness and secrecy are usually stated separately. Correctness requires that Alice’s and Bob’s output keys differ only with a small probability:

Pr⁡[KA≠KB and the protocol does not abort]≤εcor.\Pr[K_A\neq K_B\ \text{and the protocol does not abort}] \leq \varepsilon_{\mathrm{cor}}.

For secrecy, the real joint state of the output key and Eve’s information is compared with an ideal state containing a uniform key independent of Eve. A common formulation is

ppass 12∥ρKE∣pass−τK⊗ρE∣pass∥1≤εsec,p_{\mathrm{pass}}\, \frac12 \left\| \rho_{KE\mid\mathrm{pass}} - \tau_K\otimes\rho_{E\mid\mathrm{pass}} \right\|_1 \leq \varepsilon_{\mathrm{sec}},

where ppassp_{\mathrm{pass}} is the probability that the protocol does not abort, τK\tau_K is the uniform-key state, and the trace norm measures distinguishability.

This formulation supports composability: the key can replace an ideal secret key inside a larger cryptographic system with a bounded change in security. In appropriately formulated compositions, failure parameters can be added conservatively. A system cannot, however, ignore the cumulative contribution of repeated sessions or the assumptions of the application consuming the key. The security parameter is part of an end-to-end accounting, not a label that makes every later operation secure.

Finite-size penalties become especially important on lossy links. Even when the transmitted pulse count is large, the number of useful detections or test-basis outcomes may be small. Increasing the preference for the key-generating basis improves sifting efficiency but reduces statistical evidence from the test basis. Similarly, dividing the data among several source intensities improves source characterization while leaving fewer events in each category.

The operational result can be a zero or negative calculated key length even when an asymptotic formula predicts a positive rate. In that case the protocol must abort rather than output a key with the claimed security. Reported performance therefore needs a block size, security parameters, and a statement of whether the result is asymptotic or finite-key.

Optical implementations and weak coherent pulses

An ideal BB84 source emits exactly one photon in each selected qubit state. Practical single-photon sources are challenging, and many systems instead use heavily attenuated laser pulses. These weak coherent pulses are convenient, fast, and compatible with telecommunications components, but they do not contain a fixed number of photons. For a phase-randomized coherent pulse with mean photon number μ\mu, the photon-number distribution is

Pμ(n)=e−μμnn!.P_\mu(n)=e^{-\mu}\frac{\mu^n}{n!}.

Some pulses contain no photons, some contain one, and a nonzero fraction contain two or more.

The multiphoton probability is

Pμ(n≥2)=1−e−μ(1+μ),P_\mu(n\geq2)=1-e^{-\mu}(1+\mu),

which is approximately μ2/2\mu^2/2 when μ\mu is small. At μ=0.1\mu=0.1, roughly 0.47%0.47\% of all emitted pulses contain at least two photons. Although this looks small, it can be security-relevant when channel loss removes most single-photon signals. An adversary may selectively exploit multiphoton pulses while making the resulting detection rate appear compatible with ordinary attenuation.

In a photon-number-splitting attack, Eve conceptually identifies multiphoton emissions without disturbing the encoded degree of freedom, retains one photon, and forwards the remainder. After basis announcement, she can measure her retained photon in the correct basis. This need not introduce the disturbance associated with intercept–resend. Consequently, a low QBER alone does not establish security for a weak-pulse implementation. The source’s photon-number statistics must enter the proof.

Optical encoding introduces additional engineering choices. Polarization encoding is conceptually direct and well suited to some free-space links, but optical fibre can transform polarization in a time-dependent manner. Phase encoding uses interferometers and relative phase differences, often between two time bins. Time-bin and phase implementations require stable interference and careful characterization of unwanted optical modes. A logical qubit must not inadvertently carry basis or bit information in pulse timing, wavelength, or another degree of freedom.

Receivers may choose their measurement basis actively with a modulator or passively through optical splitting. Detection commonly uses semiconductor avalanche photodiodes or superconducting nanowire detectors. Important parameters include efficiency, timing jitter, dark counts, recovery time, and behavior under unusually bright illumination. A detector with excellent nominal efficiency is not necessarily described by the same security model as another detector with a different response.

Random choices are equally physical. Bit values, bases, and any decoy intensities must have the independence and unpredictability assumed in the analysis. A pseudorandom number generator can provide computationally useful randomness, but its output does not automatically satisfy an information-theoretic assumption against an adversary with unlimited computation. The implementation and security claim must specify how random choices are generated and protected.

Decoy states and practical key-rate estimation

Decoy-state quantum key distribution was developed to make weak-pulse implementations secure and efficient despite multiphoton emissions. Won-Young Hwang proposed a decoy-state method in 2003. In 2005, Hoi-Kwong Lo, Xiongfeng Ma, and Kai Chen, and independently Xiang-Bin Wang, developed influential practical approaches. The method preserves BB84’s qubit-state encoding while adding randomly selected optical intensities that help characterize the channel’s response to different photon-number components.

Alice intersperses signal pulses with one or more decoy intensities. Under the standard model, pulses with the same photon number and encoded state are otherwise indistinguishable, regardless of which intensity setting produced them. Eve can distinguish photon number in principle, but should not have an additional label identifying the intensity choice for a given photon-number component. Alice announces the intensity settings after transmission so that detection statistics can be separated into classes.

Let YnY_n denote the probability of an accepted detection conditional on emitting nn photons. For an intensity μ\mu, the observed gain, or detection probability per emitted pulse, satisfies

Qμ=∑n=0∞Pμ(n)Yn.Q_\mu = \sum_{n=0}^{\infty}P_\mu(n)Y_n.

Similarly, if ene_n is the error rate associated with the nn-photon component,

EμQμ=∑n=0∞Pμ(n)Ynen.E_\mu Q_\mu = \sum_{n=0}^{\infty}P_\mu(n)Y_ne_n.

Measurements at several intensities constrain these mixtures and allow bounds on the single-photon yield Y1Y_1 and its error contribution. Vacuum or near-vacuum settings also help estimate background detections.

A representative asymptotic key-rate expression is

R≥q[−QμfECh2(Eμ)+Q1(1−h2(e1ph))],R\geq q\left[ -Q_\mu f_{\mathrm{EC}}h_2(E_\mu) + Q_1\bigl(1-h_2(e_1^{\mathrm{ph}})\bigr) \right],

where qq is a basis-selection factor, Q1=Pμ(1)Y1Q_1=P_\mu(1)Y_1 is the single-photon contribution to detections, and e1phe_1^{\mathrm{ph}} is a bound on the single-photon phase-error rate. The negative term accounts for reconciliation of the detected signal data; the positive term represents extractable secrecy from the single-photon component. Finite-key versions replace ideal parameters with statistically justified bounds and add security overheads.

Decoy states do not make multiphoton pulses behave like single photons. Instead, they prevent an adversary from hiding unrestricted manipulation of the photon-number distribution behind channel loss. Under common idealized assumptions, the achievable rate can scale approximately linearly with channel transmittance, rather than suffering the poorer scaling of basic weak-pulse methods without decoy analysis.

The indistinguishability assumption is crucial. Intensity modulation can unintentionally change pulse shape, spectrum, timing, or phase properties, creating information that distinguishes signal and decoy settings. Likewise, the Poisson-mixture description normally requires suitable phase randomization. A valid implementation must connect its measured optical behavior to the assumptions used in the parameter-estimation equations.

Side channels, device attacks, and alternative architectures

BB84 proofs apply to specified physical models, not to a protocol name printed on a product. An implementation may reproduce the expected basis statistics while leaking information through an unmodeled channel. A side-channel attack exploits such behavior rather than contradicting the mathematical security proof. Possible leakage includes electrical emissions, optical back-reflections, distinguishable pulse shapes, or detector responses that depend on parameters outside the intended qubit space.

Detector attacks have been especially influential. Efficiency mismatch can make detection probabilities depend on the arrival time or another variable controlled by Eve. Time-shift attacks exploit such differences to bias which detector is likely to register a signal. Detector-blinding demonstrations showed that bright illumination could, for some avalanche-photodiode receivers, alter the detector’s operating regime and allow controlled responses to later light pulses. Such attacks concern the behavior of specific devices; they do not imply that every receiver responds identically.

Transmitters also present attack surfaces. In Trojan-horse attacks, Eve injects light into an apparatus and studies returning reflections for information about internal settings. Optical isolation, filtering, monitoring, and source characterization can help bring the apparatus closer to a justified model, but individual countermeasures have defined operating ranges. A security claim requires evidence that relevant deviations are either bounded by the proof or excluded by the stated assumptions.

One architectural response is measurement-device-independent quantum key distribution, proposed in 2012. Alice and Bob each prepare states and send them to a central station that performs a joint measurement and announces its result. The station need not be trusted for secrecy. Security is designed to remain valid even when its measurement apparatus is controlled by an adversary, thereby removing detector side channels at that station from the trusted-device boundary.

This does not eliminate all device assumptions. The users’ state preparation, randomness, isolation, and classical processing still require suitable characterization. Measurement-device-independent operation also introduces engineering demands, notably interference between independently prepared optical signals and precise control of their distinguishability. Its network topology differs from ordinary one-way BB84, so key rates and resource requirements must be compared on a consistent basis.

Device-independent quantum key distribution goes further by deriving security from observed nonclassical correlations, usually involving a suitable Bell-inequality violation. It requires demanding experimental conditions and is not obtained merely by adding an entanglement source to a conventional BB84 receiver. Conventional BB84, measurement-device-independent protocols, and device-independent protocols make different trade-offs between trusted hardware assumptions and achievable performance.

Transmission limits and use in communication systems

The most direct obstacle to long-distance BB84 is optical attenuation. For a fibre with attenuation coefficient α\alpha, measured in decibels per kilometre, the channel transmittance over length LL is

ηch=10−αL/10.\eta_{\mathrm{ch}}=10^{-\alpha L/10}.

Using α=0.2 dB/km\alpha=0.2\ \mathrm{dB/km} as a representative value near the telecommunications window, 100 kilometres of fibre transmits about 1%1\% of the input light, and 200 kilometres about 0.01%0.01\%. Coupling losses, filters, and detector inefficiency further reduce the observed detection probability. These figures concern channel loss, not final secret-key throughput.

At high loss, dark counts and background events become a larger fraction of the detections. They increase the error rate and weaken parameter estimates, eventually preventing a positive key length. Longer acquisition can improve statistical confidence but cannot overcome every physical limitation. Reported performance should distinguish the optical pulse rate, raw detection rate, sifted rate, and final secret-key rate. Only the last includes all privacy and security deductions.

A conventional optical amplifier cannot simply restore an arbitrary unknown qubit without introducing noise. This is a fundamental difference from ordinary classical communications, where signals can be detected, regenerated, and retransmitted along a route. A quantum repeater seeks to extend quantum communication using resources such as stored entanglement, entanglement swapping, and error management rather than straightforward signal copying. Such systems require capabilities beyond those of a standard point-to-point BB84 link.

A practical alternative is a chain of trusted key-relay nodes. Adjacent nodes establish separate keys and use protected classical messages to relay key material across the network. This can extend geographic coverage, but secrecy then depends on the intermediate nodes remaining trusted under the relay architecture. Free-space and satellite links provide another approach by changing the transmission geometry. Their performance depends on diffraction, pointing, atmospheric conditions, background light, and available contact times.

BB84’s role must also be distinguished from post-quantum cryptography. A sufficiently capable quantum computer running Shor’s algorithm would undermine widely used factoring- and discrete-logarithm-based public-key systems. Post-quantum algorithms seek replacements that operate over ordinary classical networks. BB84 instead requires quantum communication hardware and offers key establishment under physical and statistical assumptions.

In deployed communication systems, the generated key must be delivered to authenticated applications, stored safely, assigned to the correct session, and erased when no longer needed. Compromise of an endpoint can expose plaintext or keys regardless of how securely the key crossed the optical channel. Integration with encryption equipment, classical authentication, network management, and key-lifecycle controls therefore remains part of the system’s security boundary.