Dual-Key Symmetric Encryption: Overview

• Dual-key symmetric encryption is a technique where two separate keys encrypt different data components, enabling partitioned confidentiality, especially in voting systems.
• It typically employs QKD-based key establishment and XOR operations to balance efficiency with cryptographic separation of vote content and voter identity.
• While not standardized, its implementations demonstrate rapid processing with architectural trade-offs in integrity and formal security guarantees.

Dual-key symmetric encryption is not a standardized cryptographic primitive with a single formal meaning across the literature. In one recent and highly specific usage, it denotes an application-layer design in which two different symmetric keys encrypt two different fields of a ballot record—one key for the vote and a second key for the voter identity—so that tallying can proceed on decrypted vote contents while identities remain encrypted (Mahmoud et al., 3 Oct 2025). Other papers use adjacent terminology for markedly different constructions, including QKD-to-DEM hybrid composition, threshold or distributed symmetric-key encryption, multi-key transform workflows in symmetric homomorphic encryption, split-share XOR masking, and optional two-password keystream generation (Dey et al., 14 Jan 2025, Mouël et al., 2024, Sharma, 2013, Adak, 2013, Lemaire, 2019). The term therefore requires context-specific interpretation.

1. Terminological scope and taxonomy

In the voting framework of “A Quantum-Secure Voting Framework Using QKD, Dual-Key Symmetric Encryption, and Verifiable Receipts,” dual-key symmetric encryption is explicitly not presented as a standard cryptographic primitive with a formal security definition, a two-key block cipher, sequential double encryption, a split-control decryption mechanism, or a key-combination construction such as $C = E_{K_2}(E_{K_1}(M))$ or $K = K_1 \oplus K_2$. Its operational meaning is narrower: two distinct symmetric keys are each used once on different plaintext components, with the stated goal of separating ballot content from voter identity during tallying (Mahmoud et al., 3 Oct 2025).

The broader literature cited here confirms that the phrase “dual-key symmetric encryption” is not used uniformly. In “Secure Composition of Quantum Key Distribution and Symmetric Key Encryption,” the relevant architecture is a qKEM/DEM hybrid, not a native two-key symmetric primitive (Dey et al., 14 Jan 2025). “Trustless Distributed Symmetric-key Encryption” treats split authority through threshold DPRF evaluation and shared key material rather than two unrelated full keys (Mouël et al., 2024). “Fully Homomorphic Encryption Scheme with Symmetric Keys” is a single-secret-key matrix-based homomorphic scheme whose extra key-transform machinery is operationally multi-key but not dual-key in the primitive-design sense (Sharma, 2013). “Dual Layer Textual Message Cryptosystem with Randomized Sequence of Symmetric Key” uses two XOR masks on two split slides, while “Pretty Modular Symmetric Encryption” offers an optional two-password mode inside one keystream generator rather than two coequal encryption keys (Adak, 2013, Lemaire, 2019).

Interpretation	Core mechanism	Representative source
Field separation by key separation	Vote and ID encrypted under different keys	Voting framework (Mahmoud et al., 3 Oct 2025)
Hybrid two-layer keying architecture	qKEM establishes key; DEM encrypts message	qHE/qKEM (Dey et al., 14 Jan 2025)
Threshold shared-key capability	Threshold DPRF enables encryption and decryption	TDiSE (Mouël et al., 2024)
Split-share parallel masking	Two slides, each XORed with its own mask	Dual-layer message scheme (Adak, 2013)
Two-password internal state generation	Two passwords drive one keystream process	PMSE (Lemaire, 2019)
Composed-key transform workflow	Matching keys and lock/unlock transforms	Symmetric FHE scheme (Sharma, 2013)

A plausible implication is that “dual-key symmetric encryption” should be treated less as a settled primitive class than as a family of context-dependent engineering patterns and protocol architectures.

2. Field-separated encryption in the voting framework

The most concrete definition in the cited material appears in Section 3.5 of the voting framework: “Each vote and its corresponding voter ID are encrypted separately using two different keys derived from QKD. The server receives only the key required to decrypt the vote, enabling it to tally results without accessing voter identities.” The construction is therefore best described as field separation by key separation rather than two-man control over one ciphertext (Mahmoud et al., 3 Oct 2025).

The parties are the Voter and the Election Committee server. The paper first describes BB84 QKD as establishing a secure symmetric key, but the dual-key protocol summary then uses two QKD-derived symmetric keys,

$K_{\text{vote}}, \quad K_{\text{id}}.$

The manuscript does not explain whether these arise from two independent BB84 sessions, from subkey derivation, or from splitting one longer QKD output. It only states: “Generate two QKD-derived symmetric keys: $K_{\text{vote}}$ and $K_{\text{id}}$.”

The BB84 workflow is described in six steps. The voter generates a random bit string, encodes each bit in a random basis—rectilinear $(\ket{0}, \ket{1})$ or diagonal $(\ket{+}, \ket{-})$—and transmits qubits over the quantum channel. The Election Committee measures each qubit in a randomly chosen basis. Both sides then announce bases over a classical channel, discard positions where bases differ, and keep matching-basis bits as the shared key. The retained bits form the shared key, and the paper adds that each QKD session creates new keys that “are used once to keep the XOR encryption secure.”

Encryption itself is performed with bitwise XOR rather than a named block or stream cipher. The intended formulas are

$$E_{\text{vote}} = \text{vote} \oplus K_{\text{vote}}, \qquad E_{\text{id}} = \text{ID} \oplus K_{\text{id}}.$$

The voter sends the pair

$(E_{\text{vote}}, E_{\text{id}})$

to the Election Committee over the classical channel, simulated with MQTT. On the server side, tallying proceeds after decrypting only the vote field: $$\text{vote} = E_{\text{vote}} \oplus K_{\text{vote}}.$$ The pseudocode instructs the server to decrypt $K = K_1 \oplus K_2$0 using $K = K_1 \oplus K_2$1, not to decrypt $K = K_1 \oplus K_2$2 unless an audit is requested, and to store $K = K_1 \oplus K_2$3 securely for dispute resolution.

The receipt mechanism is hash-based. The intended formula is

$K = K_1 \oplus K_2$4

with server-side recomputation

$K = K_1 \oplus K_2$5

The voter computes $K = K_1 \oplus K_2$6, sends the ciphertext pair, receives $K = K_1 \oplus K_2$7, and checks whether $K = K_1 \oplus K_2$8. In operational terms, the protocol confirms receipt and storage consistency of the encrypted package rather than correctness of tallying or inclusion in a final count.

3. Security semantics, guarantees, and limits

The voting framework’s central security intuition is privacy-by-separation. If only $K = K_1 \oplus K_2$9 is available to the server, then the server can read the vote but not the associated identity, while $K_{\text{vote}}, \quad K_{\text{id}}.$0 remains encrypted. This yields what the paper implicitly treats as separation between content and identity during tallying (Mahmoud et al., 3 Oct 2025).

Under that interpretation, the natural compromise cases are partitioned rather than threshold-based. Compromise of $K_{\text{vote}}, \quad K_{\text{id}}.$1 reveals vote content but not identity, assuming $K_{\text{vote}}, \quad K_{\text{id}}.$2 remains secret. Compromise of $K_{\text{vote}}, \quad K_{\text{id}}.$3 reveals identity but not vote content, assuming $K_{\text{vote}}, \quad K_{\text{id}}.$4 remains secret. Compromise of both keys links vote and identity. The design therefore provides partitioned confidentiality, not threshold confidentiality, and both keys are not required to decrypt the vote.

The role of QKD is limited to key establishment. QKD is intended to replace classical key exchange with BB84-based quantum key distribution so that interception becomes detectable through induced errors. The manuscript simulates channel noise by introducing random error rates between 0 and 0.2 and mitigates this by retaining only bits measured in matching bases. It does not discuss privacy amplification, authentication of the classical channel, reconciliation beyond basis matching, or composable QKD security.

The integrity story is substantially weaker than the confidentiality story. There is no MAC and no authenticated encryption. XOR alone provides confidentiality only if the keying assumptions are ideal; it does not provide ciphertext integrity. The SHA-256 receipt is unkeyed, so it does not authenticate the sender or prevent active substitution by an adversary who can replace ciphertexts and recompute the hash. The receipt therefore provides only a weak form of confirmation of receipt and storage consistency. The paper also does not furnish mechanisms that actually achieve non-repudiation, coercion resistance, receipt-freeness, or strong end-to-end verifiability. Its claims of privacy, integrity, and verifiability are therefore intuitive and architectural rather than theorem-backed.

A further limitation is formalization. The paper gives no formal adversary model, no indistinguishability notion, no game-based proof, and no theorem specifying what is guaranteed if one key or one party is compromised. The repeated claim that the method “simulates the effect of homomorphic tallying” must therefore be read narrowly: the scheme does not support arithmetic on ciphertexts and does not instantiate homomorphic encryption. It only withholds decryption of the identity field during tallying.

4. Relation to hybrid, threshold, and homomorphic designs

The closest rigorous comparison in the cited set is the qKEM/DEM framework of “Secure Composition of Quantum Key Distribution and Symmetric Key Encryption.” That paper addresses the problem of composing QKD-established secret material with a symmetric encryption scheme for unrestricted-length messages. Its architecture is explicitly hybrid: $K_{\text{vote}}, \quad K_{\text{id}}.$5 and its main theorem states that if the qKEM is $K_{\text{vote}}, \quad K_{\text{id}}.$6-IND-OT secure and the DEM is $K_{\text{vote}}, \quad K_{\text{id}}.$7-IND-OT secure, then the resulting qHE is $K_{\text{vote}}, \quad K_{\text{id}}.$8-IND-OT secure (Dey et al., 14 Jan 2025). This is a two-stage keying architecture, but not a dual-key symmetric cipher in the sense used or implied by the voting framework.

A different notion appears in “Trustless Distributed Symmetric-key Encryption,” where the effective symmetric capability is shared across multiple parties via Shamir secret sharing and a DDH-based distributed PRF. In that construction, the final share held by participant $K_{\text{vote}}, \quad K_{\text{id}}.$9 is

$K_{\text{vote}}$0

and the global secret is

$K_{\text{vote}}$1

Partial evaluations

$K_{\text{vote}}$2

are combined as

$K_{\text{vote}}$3

In a $K_{\text{vote}}$4-of-$K_{\text{vote}}$5 instantiation, both parties must cooperate for both encryption and decryption, which is operationally close to dual control, but conceptually it is one logical shared key realized through threshold cryptography rather than two independent symmetric keys (Mouël et al., 2024).

The symmetric homomorphic scheme of (Sharma, 2013) further sharpens the distinction. Its base encryption is single-key: $K_{\text{vote}}$6 with decryption

$K_{\text{vote}}$7

Its later matching-key relations such as $K_{\text{vote}}$8 or $K_{\text{vote}}$9 enable transform workflows and multi-user processing, but the work itself explicitly does not define a formal dual-key symmetric encryption construction. The same caveat applies, in a different form, to the split-slide XOR system of (Adak, 2013) and the two-password PMSE design of (Lemaire, 2019): both are operationally multi-secret or multi-component, but neither is a standard dual-key primitive.

Taken together, these papers show that apparently similar terminology can refer to at least four different ideas: field-separated encryption, hybrid key-establishment-plus-DEM composition, threshold shared-key operation, and multi-component masking or transform workflows.

5. Implementation profile and reported performance

The voting framework is implemented with a concrete but sparse stack: BB84 for QKD, QASM 3 and Qiskit Aer for quantum representation and simulation, bitwise XOR for symmetric encryption, SHA-256 for receipts, and MQTT for classical message transport (Mahmoud et al., 3 Oct 2025). No AES, ChaCha20, HMAC, GCM, Poly1305, or digital signatures are used in the described design.

In the MQTT-based simulation, the voter publishes $K_{\text{id}}$0, the server consumes the pair, decrypts $K_{\text{id}}$1 using $K_{\text{id}}$2, computes the receipt hash over both ciphertexts, and returns the hash as the receipt confirmation. The paper states that MQTT “accurately model[s] the transmission conditions that a real-world voting system may encounter” and enables “rapid exchange of quantum keys and vote information,” although the quantum channel itself is simulated separately through Qiskit and QASM 3.

The reported evaluation is limited but specific. Vote processing time per vote is approximately $K_{\text{id}}$3 seconds, and estimated throughput is approximately 10 million votes in approximately 16 minutes. The testing environment is an 8-core MacBook Air with a Qiskit simulator. The QKD shot count is 10,000 shots per measurement. Among tested key sizes, 4 bits is reported as optimal for stability, while 2-bit and 32-bit key sizes are described as unstable. The summary table characterizes the encryption scheme as “Bitwise XOR using shared key from BB84” and the tallying method as “Dual-Key Symmetric encryption (efficient, less resource-intensive than full HE).”

These results should be read carefully. The paper does not isolate the overhead of the dual-key design relative to a single-key design, and it does not report separate encryption time, decryption time, hash time, or memory and storage overhead from retaining encrypted IDs. It also does not quantify the incremental MQTT latency of carrying two ciphertext fields instead of one. The main claimed trade-off is architectural: full homomorphic encryption offers stronger encrypted-domain processing but much higher complexity, whereas the dual-key separation pattern reduces overhead at the cost of much weaker cryptographic functionality.

6. Misconceptions, controversies, and stable points of interpretation

Several misconceptions recur around the term. In the voting framework, dual-key symmetric encryption does not mean a two-key block cipher, sequential double encryption of one message, a composite-key construction, or threshold decryption of one ciphertext (Mahmoud et al., 3 Oct 2025). It means separate encryption of separate fields. In that sense, the most stable description is an application-layer access-separation pattern.

A second misconception is equivalence to homomorphic encryption. The voting paper repeatedly claims that its method “simulates the effect of homomorphic tallying” or “achieves the same functional objective as homomorphic encryption.” Technically, that equivalence does not hold. The system decrypts vote contents conventionally and merely withholds the identity-decryption key during tallying. It does not permit arithmetic on ciphertexts, and it is therefore not equivalent to homomorphic encryption in cryptographic capability.

A third misconception is that any use of two secrets automatically yields a rigorous dual-key construction. The split-slide scheme of (Adak, 2013) uses two independently encrypted shares,

$K_{\text{id}}$4

but this is better understood as two-channel secret splitting plus independent masking. PMSE’s two-password mode similarly mixes two passwords into one evolving keystream state rather than defining a primitive of the form $K_{\text{id}}$5 (Lemaire, 2019). These designs are relevant analogies, but not canonical instantiations of a standard dual-key symmetric cipher.

The stable point across the cited literature is therefore negative rather than positive: no single formal definition covers all uses of “dual-key symmetric encryption.” In the most specific cited usage, the term denotes field separation by key separation in a QKD-supported voting workflow. In adjacent usages, it denotes hybrid composition, threshold shared control, serial key transforms, or parallel masking of split components. This suggests that rigorous discussion of the topic should begin by specifying which of those meanings is intended before any security claim is evaluated.