Plaintext-Delayed Chaotic Cipher (PDCC)

Updated 11 July 2026

PDCC is a chaotic encryption design that explicitly incorporates delayed plaintext inputs into its diffusion process for enhanced security.
By embedding historical plaintext alongside ciphertext feedback, PDCC mitigates vulnerabilities seen in traditional plaintext non-delayed chaotic ciphers.
The PDCC framework shifts focus from mere statistical validation to structural analysis, emphasizing state recurrence and message-specific diffusion.

Searching arXiv for recent PDCC-related papers and the cited comparative background. Plaintext-Delayed Chaotic Cipher (PDCC) denotes a chaotic-cipher design in which the diffusion model is altered so that delayed plaintext terms enter the encryption or inverse-diffusion structure alongside delayed ciphertext terms. In the formulation proposed as a secure improvement over plaintext non-delayed chaotic ciphers (PNDCCs), the defining contrast is with diffusion equations of the form $C(k)=f_D(P(k),C(k-1),K(k))$ , where plaintext appears only as the current symbol $P(k)$ . PDCC was introduced as a remedy for what was presented as a structural weakness of PNDCCs: despite good statistical indicators, representative multi-stage permutation–diffusion image ciphers remained vulnerable to several cryptanalytic attacks. The PDCC proposal therefore belongs to a broader attempt to move chaotic cryptography away from purely statistical validation and toward architectures whose state evolution depends more deeply on message history (Wang et al., 14 Sep 2025).

1. Definition and conceptual scope

The paper that explicitly introduces PDCC defines PNDCC as a chaotic cipher whose diffusion equation has the form

$C(k)=f_D\big(P(k),C(k-1),K(k)\big),$

with $P(k)$ the plaintext symbol or pixel, $C(k)$ the ciphertext symbol or pixel, $C(k-1)$ ciphertext feedback, and $K(k)$ the keystream or diffusion key. The defining limitation is that plaintext enters without delay: terms such as $P(k-1)$ , $P(k-2)$ , and further delayed plaintext variables do not appear explicitly. PDCC is proposed by modifying the diffusion inverse or model so that it contains both delayed plaintext terms and delayed ciphertext terms, and the paper’s generic inverse-diffusion stages explicitly include dependencies such as $P_1(k-1),P_1(k-2),\cdots,P_1(k-n_1)$ together with $P(k)$ 0 (Wang et al., 14 Sep 2025).

Within chaotic cryptography, this places PDCC in a distinct category. It is not merely a chaos-based keystream generator, not merely a message-masking chaotic carrier, and not simply a permutation–diffusion image cipher with ciphertext chaining. Its central claim is architectural: delayed plaintext must participate explicitly in the diffusion mechanism. This makes PDCC different from schemes in which chaos is only an entropy source, as in physical-layer XOR stream ciphers based on chaotic-light-derived true random numbers, and different from synchronous chaotic stream ciphers where plaintext never perturbs the chaotic dynamics (Zhao et al., 2010).

A common misconception is to treat any chaotic system with a delay as a PDCC. The comparative literature shows that this is too broad. Delayed optical feedback in a laser, sparse sampling of a hyperchaotic trajectory, or a one-sample shift used in random-number post-processing are delays in entropy generation or observation, not plaintext delay in the encryption architecture (Zhao et al., 2010, Vidal et al., 2013). PDCC, by contrast, is defined by delayed plaintext participation in diffusion.

2. Formal structure of delayed plaintext diffusion

The formal motivation for PDCC follows from the claim that existing chaotic-cipher diffusions “invariably” take the PNDCC form. In that model, the current ciphertext symbol is produced from the current plaintext, previous ciphertext, and a key-dependent diffusion function. The absence of explicit $P(k)$ 1 terms means that plaintext history is not structurally embedded in the diffusion equation (Wang et al., 14 Sep 2025).

PDCC changes this by introducing delayed plaintext into the inverse-diffusion or diffusion model. The supplied formulation is partial, but its structure is clear: a stage variable such as $P(k)$ 2 is expressed as a function of delayed plaintext variables $P(k)$ 3 together with current and delayed ciphertext variables. This makes plaintext history a first-class part of the recurrence, rather than a by-product of ordinary chaining (Wang et al., 14 Sep 2025).

This distinction matters because many “feedback” chaotic ciphers do not satisfy it. In message-embedded 2-D chaotic-map schemes, plaintext is mixed with chaotic state and the resulting scrambled output is said to drive the chaotic system, but the paper gives no explicit delayed plaintext injection such as $P(k)$ 4, no formal delayed feedback recurrence, and no self-synchronizing formulation (Mishra et al., 2012). Similarly, 1-D Logistic/NLFSR schemes describe feeding the scrambled output back into the chaotic system so that the chaotic dynamics are changed continuously, yet they provide no delay operator and no recurrence of the form $P(k)$ 5 (Mishra et al., 2013). These systems are plaintext-coupled or message-embedded, but not formal PDCCs.

A plausible implication is that PDCC should be understood less as a generic “cipher with feedback” than as a specific subclass in which delayed plaintext variables are explicit state inputs. This narrows the term considerably and distinguishes PDCC from several adjacent architectures often grouped together in the chaotic-cryptography literature.

3. PDCC and the critique of plaintext non-delayed chaotic ciphers

The immediate historical context of PDCC is cryptanalytic rather than purely constructive. The PDCC paper presents a “typical example” of a three-stage permutation–diffusion–permutation PNDCC and reports that, although all of its statistical indicators show good performance, the scheme can be broken using four different attacks: a differential attack based on homogeneous operations, an S-PTC attack, an impulse-step-based differential attack (ISBDA), and a chain attack (Wang et al., 14 Sep 2025).

The broad claim is that fulfillment of statistical criteria is not a sufficient condition for the security of PNDCC. The paper further states that, from a mathematical model of multi-stage PNDCC, the proposed chain attack can successfully break a class of multi-stage PNDCCs. Its key technique depends on revealing all permutations by reconstructing the same decryption chain and then breaking the entire diffusion process by solving a system of simultaneous equations (Wang et al., 14 Sep 2025).

This critique aligns with earlier cryptanalytic findings on chaotic image ciphers. A modified substitution–diffusion cipher based on chaotic standard and logistic maps was shown to collapse to

$P(k)$ 6

so that one known or chosen plaintext–ciphertext pair suffices to recover an equivalent key for all same-size images under the same master key (0912.3050). A permutation-plus-XOR masking design using Arnold cat map shuffling and a Chen-system keystream was likewise broken with two chosen plaintext-ciphertext pairs because the cipher decomposed as

$P(k)$ 7

making the permutation and keystream independently recoverable (Mukherjee et al., 2018).

The relevance to PDCC is structural. These attacks do not show merely that particular chaotic maps are weak; they show that separable architectures, plaintext-independent masks, and affine diffusion can remain vulnerable even when conventional image-encryption statistics look satisfactory. PDCC is presented as a response to that diagnosis.

4. Relation to neighboring chaotic-cipher families

The literature surrounding PDCC includes several neighboring architectures that are close enough to be informative yet different enough to require separation.

Family	Characteristic mechanism	Relation to PDCC
PNDCC	$P(k)$ 8	Baseline family PDCC seeks to replace
Message-embedded chaotic ciphers	Scrambled output drives future chaotic evolution	Adjacent, but usually no explicit delayed plaintext equation
Chaos-based XOR stream ciphers	Plaintext XORed with chaotic keystream or TRN stream	Not PDCC
Delayed chaotic masking schemes	Plaintext added to delayed chaotic carrier	PDCC-like in masking, but not necessarily plaintext-driven state update

Message-embedded schemes based on Arnold’s cat map and Duffing’s map are conceptually adjacent because plaintext is mixed with chaotic state and the scrambled output is used to drive the chaotic system, producing plaintext-coupled state evolution. Yet the exact feedback law, delay length, and synchronization model are not formalized, so their relation to PDCC is conceptual rather than exact (Mishra et al., 2012). Logistic/NLFSR text ciphers show a similar pattern: the scrambled output is fed back into the chaotic system, but no explicit delayed plaintext recurrence is given, and known-plaintext robustness is limited in the simpler variants (Mishra et al., 2013).

By contrast, some schemes are clearly outside the PDCC category. High-speed physical-layer stream ciphers based on chaotic light use chaotic semiconductor lasers as entropy generators for true random keys and then compute ciphertext by bitwise XOR; there is no delayed plaintext feedback, no delayed plaintext-dependent state update, and no synchronized chaotic carrier directly embedding the message (Zhao et al., 2010). A smartphone-oriented hyperchaotic stream cipher similarly samples a hyperchaotic trajectory, binarizes it, and XORs plaintext with the resulting keystream; plaintext does not perturb the ODE state, and there is no plaintext-delayed or ciphertext-feedback mechanism (Vidal et al., 2013).

There are also PDCC-adjacent designs that incorporate delayed or history-dependent masking without adopting the PDCC name. A reset-based synchronized multidelay masking scheme forms

$P(k)$ 9

using multiple delayed chaotic states in a separate message channel, while synchronization is handled on an intermittent second channel (Ambika et al., 2010). A dynamic DNA image cipher generates a “dynamic one-time pixel”

$C(k)=f_D\big(P(k),C(k-1),K(k)\big),$ 0

where $C(k)=f_D\big(P(k),C(k-1),K(k)\big),$ 1 depends on future plaintext pixels and $C(k)=f_D\big(P(k),C(k-1),K(k)\big),$ 2 depends on previous ciphertext pixels, and then adds previous-cipher DNA in the core recurrence. This gives strong plaintext dependence and ciphertext feedback, though the chaotic map itself is still key-driven rather than message-driven (Patidar et al., 2022).

5. Cryptanalytic significance of delay, feedback, and completeness

The existence of delayed dependence does not by itself guarantee cryptographic strength. This is shown especially clearly by delayed-dynamics block ciphers such as EDDy and ExEDDy. EDDy updates bits by delayed permutation coupling,

$C(k)=f_D\big(P(k),C(k-1),K(k)\big),$ 3

and exhibits poor avalanche behavior: the average Hamming-distance response to a one-bit plaintext flip scales as $C(k)=f_D\big(P(k),C(k-1),K(k)\big),$ 4, because the permutation decomposes into independent cycles and perturbations remain confined within a cycle. ExEDDy retains the delayed-permutation term but multiplies each update by a global sign depending on the sum of all bits, thereby introducing nonlinear all-to-all coupling and bringing avalanche close to the ideal $C(k)=f_D\big(P(k),C(k-1),K(k)\big),$ 5 (Melchert, 2012).

This comparison has direct relevance to PDCC. It shows that delay alone is insufficient; what matters is whether delayed coupling produces completeness, strong nonlinear mixing, and broad perturbation spreading. A plausible implication is that formal PDCC designs must be evaluated not only for the presence of delayed plaintext terms, but also for the connectivity and nonlinearity of the induced dependence graph.

The same caution appears in modern cryptanalysis of chaotic ciphers by learned inversion. A deep-learning-based known-plaintext attack trains a neural network directly on plaintext–ciphertext pairs and uses the trained model as an equivalent decryption system, without recovering the secret key. The experiments are on simple chaotic image ciphers with fixed parameters, not on PDCC specifically, but the central lesson is clear: if encryption under a fixed key behaves like a deterministic, learnable mapping, then supervised models may approximate the inverse sufficiently well to decrypt future samples from the same configuration (Wang et al., 2021). This suggests that PDCC security cannot rest solely on complexity of equations or the use of delay terms; instance-specific learnability is also relevant.

6. Design implications, scope, and open questions

PDCC was proposed as a secure improvement over PNDCC, but its broader significance lies in how it reframes chaotic-cipher design. First, it treats delayed plaintext as a structural ingredient rather than an incidental consequence of chaining. Second, it directs attention away from histogram, correlation, and entropy tests as primary evidence of security and toward recurrence structure, equivalent-key collapse, permutation recoverability, and diffusion-chain reconstruction (Wang et al., 14 Sep 2025).

The comparative literature also clarifies what PDCC is not. It is not any cipher that uses a delayed chaotic oscillator, not any chaotic stream cipher with XOR masking, and not any message-embedded chaotic map whose feedback law is unspecified. Those designs may be related, but the defining feature of PDCC is explicit delayed plaintext participation in the diffusion architecture (Zhao et al., 2010, Mishra et al., 2012).

At the same time, the literature indicates that PDCC should not be idealized. Earlier message-coupled chaotic ciphers often omit exact encryption and decryption equations, synchronization analysis, finite-precision treatment, or rigorous attack models (Mishra et al., 2012, Mishra et al., 2013). This suggests that the long-term value of PDCC will depend not only on the delayed-plaintext idea itself but also on the precision with which its state equations, implementation assumptions, and cryptanalytic boundaries are specified.

In that sense, PDCC marks a shift in emphasis within chaotic cryptography. The key proposition is not simply that “delay” improves security, but that the absence of plaintext delay in widely used PNDCC diffusion structures is itself a structural vulnerability. Whether that proposition yields a robust design framework depends on demonstrating that delayed plaintext terms produce non-separable, non-affine, message-specific diffusion without collapsing under differential, chain, equivalent-key, or learned known-plaintext attacks (Wang et al., 14 Sep 2025, Wang et al., 2021).