{"id":"61016d4e-23a1-44a1-abbb-ffad7b60bf61","arxiv_id":"2505.00593","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"A feature-aware chaotic image encryption scheme using Sobel-edge segmentation, logistic-map permutation, and hash-chained XOR confusion is presented and evaluated on three test images.","lead":"This paper proposes a chaotic image encryption scheme that separates high-detail and smooth pixels before scrambling and encrypting image blocks. It is worth reading as an example of IoT-oriented encryption design, but the absence of decryption and performance measurements leaves the main claims unverified.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stage-1 FAPS is plaintext-derived and not invertible from ciphertext; without keyed membership/order metadata, no decryption exists, so the scheme cannot function as an encryption system.","rationale":"The Reader identified the same core weakness: the receiver has no specified way to invert the plaintext-dependent Feature-Aware Pixel Segmentation. This is the most load-bearing concern because an encryption scheme with no defined decryption is not an encryption scheme. The paper's own Algorithm 1 describes only encryption, and the conclusion claims suitability for real-time IoT deployment without any decryption procedure, efficiency measurements, or complexity analysis. The missing decryption is not a stylistic omission; it is a correctness failure. Even if all reported security metrics were reproduced exactly, they would describe a one-way transformation rather than a cipher. The absence of runtime or memory benchmarks is a separate unsupported claim, but the invertibility gap is logically prior: no amount of performance data can make a non-invertible transformation usable for confidential communication. The reader's verdict of REJECT is therefore appropriate and should not be changed. The concrete test above would settle the concern by forcing the authors to exhibit an inverse for Stage 1 or acknowledge that plaintext-sized metadata is required.","tokens_in":6919,"tokens_out":2650,"duration_ms":30677,"concrete_test":"Implement the full encryption pipeline of Eqs. (1)-(26) on a 256x256 test image, then attempt to write a decryption function that takes only the ciphertext and the keyed initial values (X0, r) and recovers the original image by reversing Stage 3, Stage 2, and Stage 1. For Stage 1, specify the inverse of Eq. (10): how the receiver determines which ciphertext pixels belong to the upper high-edge half and which to the lower low-edge half, and how original coordinates are recovered. If the only way to do this is to include the segmentation mask or sorted-order indices as auxiliary plaintext-dependent data, then the scheme is not a standalone encryption system. A successful test would produce a complete, executable decryptor that reconstructs the plaintext without using any plaintext-derived side information beyond the initial key.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is an encryption scheme, but the paper only specifies a forward direction. In Section II-A, Eqs. (6)-(10), pixels are assigned to high-edge and low-edge sets using Sobel magnitudes and an Otsu threshold computed from the plaintext. The pixels are then sorted descending/ascending and concatenated into a new image I'. To decrypt, a receiver must know, for every pixel in the ciphertext, whether it came from P'_HE or P'_LE and what its original coordinates were. The paper defines no keyed metadata, no permutation map, and no auxiliary information that would let Stage 1 be inverted. The only secret inputs are the password-derived SHA-256 values and logistic-map parameters (Section II-B, II-C), which determine block permutations and XOR seeds, not the plaintext-derived segmentation. Since edge classification and Otsu thresholding depend on all plaintext intensities, the receiver cannot recompute them from the ciphertext without already knowing the plaintext. If one tried to transmit the segmentation mask or sorted-order indices, the auxiliary data would be on the order of the image size, contradicting the claimed lightweight design and effectively encrypting only the pixel values while leaking their structural grouping. This is not a missing optimization or missing benchmark; it is an internal inconsistency in the described protocol. The reported correlation and entropy values, even if accurate, do not establish encryption because the operation is not shown to be reversible. The differential-attack claim in Section III-D is likewise weakened: a one-bit plaintext change alters the Otsu threshold and the segmentation itself, so the reported avalanche says nothing about a fixed algorithm with a fixed key. The load-bearing assumption is that Stage 1 can be reversed, and the manuscript neither states nor implements that inverse.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a three-stage chaotic image encryption pipeline for IoT and edge networks: Feature-Aware Pixel Segmentation (FAPS), Chaotic Chain Permutation, and Chaotic Chain Confusion. The forward direction is presented with equations and a pseudocode algorithm. The authors report near-zero correlation coefficients and entropy values close to 7.998 for three test images, and they claim resistance to differential attacks and suitability for resource-constrained environments. However, the manuscript does not specify any decryption procedure, does not explain how the receiver can invert the plaintext-dependent FAPS stage, does not define the key derivation from the password, and provides no quantitative differential, key-sensitivity, complexity, or comparison measurements. As a result, the central claims of the abstract and conclusion are not supported by the evidence.","tokens_in":7167,"tokens_out":4829,"duration_ms":51632,"significance":"If the scheme were complete and validated, the idea of content-aware pixel segmentation to break correlation before chaotic permutation could be of interest for lightweight image encryption in edge settings, and the paper does provide a clearly written forward-direction pipeline with pseudocode. However, the proposed process is not shown to be an encryption scheme because decryption is undefined, and the reported entropy and correlation figures do not address the essential cryptographic requirements. The paper also credits its prior Permutex work on feature-extraction-based permutation but does not include a comparative evaluation against that or any other baseline. The significance of the contribution, as presently demonstrated, is therefore low.","major_comments":[{"comment":"The scheme as specified has no decryption procedure. In Stage 1, pixels are assigned to high-edge and low-edge sets using Sobel magnitudes and an Otsu threshold computed from the plaintext (Eqs. (5)-(7)), then sorted and rearranged into a new image (Eqs. (8)-(10)). The receiver cannot invert this stage because the group membership and original pixel coordinates are not derivable from the ciphertext or from any keyed information defined in Stages 2 or 3. The manuscript never specifies keyed metadata, a permutation map, or an auxiliary mask that would allow the segmentation to be reversed. Without an inversion of Stage 1, the entire method is a forward-only transform rather than an encryption scheme.","section":"Section II-A and Algorithm 1"},{"comment":"The Abstract and Conclusion claim that the scheme 'resists differential cryptographic attacks' and exhibits a high avalanche effect, but Section III-D reports only visual difference images after a one-bit plaintext change. No NPCR, UACI, or any quantitative avalanche metric is provided, and no key-sensitivity or key-space analysis is reported. The claimed resistance to differential attacks is therefore unsupported by the presented evidence.","section":"Section III-D and Abstract"},{"comment":"The conclusion states that the method is 'lightweight yet robust' for resource-constrained environments, but the paper reports no execution time, memory footprint, throughput, energy consumption, or asymptotic complexity measurements. In particular, the scheme performs a SHA-256 hash per 32x32 block in the permutation stage and per 16x16 block in the confusion stage, which introduces a substantial computational cost that is not evaluated. No comparison with AES, RSA, or existing chaos-based image ciphers is provided, so the lightweight claim is not demonstrated.","section":"Section IV (Conclusion) and Section III"},{"comment":"The key setup is undefined. Figure 2 shows a password being hashed with SHA-256 to generate control parameters, but the text states that the initial condition X0 is 'chosen randomly within the chaotic range' and gives no equation or procedure linking the password-derived hash to X0 and r for either the permutation stage or the initial seed matrix S1. A receiver cannot reproduce the keystream without this mapping. Equations (14), (21), and (25) only describe updates from previous block hashes; they do not specify how the shared secret initializes the chaos.","section":"Sections II-B and II-C, Eqs. (14) and (21)"}],"minor_comments":[{"comment":"The heading appears as 'RESULTS ANDSECURITYANALYSIS' with a missing space, and Section I states that security analysis is in Section IV, but it is actually presented in Section III.","section":"Section III heading"},{"comment":"The third stage in the block diagram is labeled 'Stage 2: Chaotic Chain Confusion'; it should be labeled Stage 3 to match the text.","section":"Figure 2"},{"comment":"The sentence 'Each clock B_i is permuted using a new permutation key' contains a typo; it should read 'Each block B_i is permuted using a new permutation key.'","section":"Section II-B, item 3"},{"comment":"Table I lists only the cipher-image correlation coefficients. The text claims that plain images exhibit strong correlations, but no plaintext correlation values are tabulated, so the claimed reduction from plaintext to ciphertext cannot be verified from the table alone.","section":"Table I"},{"comment":"Several references, including [6], [10], and [12], lack DOIs or stable URLs, which makes independent verification difficult.","section":"References"}],"recommendation":"reject","confidential_remarks":"The fundamental problem is that the algorithm is not an encryption scheme because decryption is undefined and the plaintext-dependent segmentation cannot be inverted without auxiliary information. This is not a missing benchmark or a mere presentation issue; it requires a redesign of the algorithm and a full decryption specification. The paper also relies on the authors' prior Permutex concept without a comparative evaluation, which raises concerns about incremental novelty, but the decisive issue remains the absent decryptor and undefined key derivation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe bottom line is that this paper describes a transform, not an encryption scheme. Stage 1, Feature-Aware Pixel Segmentation, sorts pixels into high- and low-edge groups based on Sobel edge magnitudes and an Otsu threshold computed from the plaintext. The receiver has no way to know which ciphertext pixels came from which group or what their original coordinates were. No keyed permutation map or segmentation mask is defined. Without that, the scheme cannot be inverted, so there is no decryption. That is a load-bearing, central flaw.\n\nWhat the paper does well: the hash-chaining idea—using SHA-256 of each processed block to update the logistic map's parameters for the next block—is reasonable and gives good plaintext sensitivity. The three-stage combination is not a direct copy of a single prior paper. The reported entropy values (7.998) and correlation coefficients (near zero) on three standard test images are consistent with a well-mixed cipher, and the numbers are at least presented.\n\nThe soft spots go beyond the missing inverse. There is no decryption algorithm, no NPCR/UACI metrics, no key-sensitivity test, no complexity analysis, no runtime or throughput measurement, and no comparison to AES or to existing chaos-based ciphers. The 'lightweight yet robust' claim rests on no efficiency data at all. The differential attack section shows a single image pair and claims avalanche without quantifying it. These omissions make the security and performance claims unsupported.\n\nThe citation pattern looks fine. The authors build on their own Permutex work, which is natural and clearly relevant.\n\nWho should read this? Possibly someone studying how hash-chained keystreams can create plaintext dependence in chaotic ciphers. But nobody should use this in its current form. The missing decryption routine alone means the paper cannot stand as an encryption proposal.\n\nRecommendation: desk reject. If the authors supply a decryption method—for example, by encrypting the segmentation mask and transmitting it as part of the ciphertext—the scheme might become a real cipher. As written, it is not.","headline":"The FAPS stage is plaintext-derived and non-invertible from the ciphertext, so the scheme cannot be decrypted; all other issues are secondary.","tokens_in":7746,"tokens_out":4728,"would_cite":false,"duration_ms":43778,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims a three-stage chaotic image encryption scheme—edge-aware pixel sorting, chain permutation, and chain XOR confusion—can push pixel correlation near zero and entropy near 8 for IoT and edge devices.","keywords":["image encryption","IoT security","edge computing","chaotic map","logistic map","feature-aware pixel segmentation","confusion-permutation","SHA-256"],"falsifier":"Encrypt a known grayscale image with the described algorithm, then attempt to decrypt it using only the cipher image and the password by inverting Stage 3, Stage 2, and Stage 1; if the original image cannot be recovered, the scheme is not a fully specified encryption system.","tokens_in":6741,"feed_emoji":"🔐","tokens_out":3818,"duration_ms":40811,"temperature":0.7,"pith_summary":"The paper proposes an image encryption scheme aimed at resource-limited IoT and edge devices, claiming it disrupts pixel correlation almost completely, reaches entropy close to 8, and resists differential attacks. The scheme's novelty is to use the image's own edge structure, found by Sobel edge detection and Otsu thresholding, to reorganize pixels before chaos-based permutation and confusion. If the claims hold, the scheme would offer a lightweight, content-aware alternative to computationally expensive methods for protecting image data in distributed intelligent systems.","feed_headline":"Edge-aware chaotic encryption aims at near-zero pixel correlation","feed_subtitle":"Sobel-detected edges split the image, then chain permutation and XOR confusion scramble blocks for IoT-class devices.","key_machinery":"The central mechanism is the trio of Feature-Aware Pixel Segmentation (FAPS), Chaotic Chain Permutation, and Chaotic Chain Confusion. FAPS uses Sobel edge detection and Otsu's threshold to split pixels into high-edge and low-edge sets, sorts them descending and ascending, and places them in the upper and lower image halves. The two chain stages then use a logistic map, with SHA-256 hashes of previously processed blocks dynamically updating the map's initial state and control parameter for each next block. This chaining makes every permutation key and confusion seed depend on all earlier processed blocks, propagating any plaintext change through the cipher image.","core_discovery":"The central claim is that combining Feature-Aware Pixel Segmentation with Chaotic Chain Permutation and Chaotic Chain Confusion yields a secure and efficient image cipher for constrained environments. Stage 1 extracts high-edge and low-edge pixels using Sobel edge detection and Otsu thresholding, then sorts them in opposite orders and stacks them in separate halves of the image. Stage 2 divides the image into 32x32 blocks and permutes each block with a logistic map whose parameters are refreshed from the SHA-256 hash of the previous permuted block. Stage 3 divides the permuted image into 16x16 blocks and XORs each block with a chaotic seed matrix that is regenerated from the hash of the previous confused block. The paper reports near-zero horizontal, vertical, and diagonal correlation coefficients, cipher-image entropy values of 7.998 on three test images, and a strong avalanche effect under a one-bit plaintext change.","pith_inferences":["An unstated corollary is that the chain-hash structure resembles a block-cipher mode of operation; the next natural test is whether it inherits classic mode weaknesses such as dependence on block boundaries or chosen-block manipulation.","The paper does not specify how a receiver would invert Stage 1; a testable extension is to define a compact, keyed record of the high/low edge mask and sort indices so that decryption is well-defined without leaking image content.","The 'lightweight' claim could be made quantitative by benchmarking the scheme against standard lightweight ciphers on a microcontroller-class device, since the paper reports no runtime, memory, or energy measurements.","The plaintext-dependent segmentation suggests an attack surface worth probing: if an adversary can influence the edge structure of a submitted image, the resulting segmentation might leak information about the grouping or key schedule."],"forward_implications":["If the reported correlation coefficients (roughly $10^{-4}$ to $10^{-3}$ in horizontal, vertical, and diagonal directions) are reproducible, statistical attacks exploiting neighbor redundancy would find almost no correlation to leverage.","If cipher images consistently reach entropy $7.998$, the output distribution is close to uniform across all 256 gray levels, making histogram-based frequency attacks ineffective.","Because each block's permutation key or confusion seed is updated from the SHA-256 hash of the previous block's output, a one-bit change in an early block should avalanche through all subsequent blocks.","Since Stage 1 derives the pixel grouping from the plaintext's edge structure, the security of the scheme is coupled to image content in addition to the secret key, which is a direct consequence of the proposed design."],"supporting_citations":[{"why":"Supplies the logistic chaotic map used for permutation keys and confusion seed matrices.","marker":"[16]"},{"why":"Introduces feature-extraction-based permutation, the line of work this scheme extends with FAPS.","marker":"[11]"},{"why":"Provides a prior chaos-based single-round S-box encryption baseline aimed at highly auto-correlated image data.","marker":"[8]"},{"why":"Represents an existing chaos-based image encryption approach for cloud storage that motivates the proposed lightweight design.","marker":"[10]"},{"why":"Shows a chaos-based encryption strengthening technique for IoT environments, supporting the use of chaotic maps in this setting.","marker":"[15]"},{"why":"Establishes the security and privacy challenges in distributed IoT networks that the paper uses to motivate the scheme.","marker":"[4]"}],"fun_headline_variants":["Edge-aware chaotic encryption nears zero pixel correlation","Chaotic chain cipher with edge segmentation for IoT images","Sobel-detected features drive chaotic image scrambling","Edge-aware chaotic cipher hits near-zero correlation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The receiver can reverse Stage 1 and recover the original pixel order from the key alone, even though the high/low edge grouping and sorting are derived from the plaintext and the paper gives no keyed description of that grouping.","fun_headline_variants_meta":{"raw":{"variants":["Edge-aware chaotic encryption nears zero pixel correlation","Chaotic chain cipher with edge segmentation for IoT images","Sobel-detected features drive chaotic image scrambling","Edge-aware chaotic cipher hits near-zero correlation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00059,"raw_usage":{"total_tokens":2775,"prompt_tokens":957,"completion_tokens":1818,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":573,"completion_tokens_details":{"reasoning_tokens":1759}},"tokens_in":573,"tokens_out":1818,"duration_ms":13931,"temperature":1.0,"reasoning_tokens":1759,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:38:45.451155+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Encrypt a known grayscale image with the described algorithm, then attempt to decrypt it using only the cipher image and the password by inverting Stage 3, Stage 2, and Stage 1; if the original image cannot be recovered, the scheme is not a fully specified encryption system.","supporting_citations":[{"cited_title":"Quantum image encryption based on baker map and 2d logistic map,","cited_arxiv_id":null,"evidence_quote":"Supplies the logistic chaotic map used for permutation keys and confusion seed matrices."},{"cited_title":"Permutex: Feature-extraction-based permutation — a new diffusion scheme for image encryption algorithms,","cited_arxiv_id":null,"evidence_quote":"Introduces feature-extraction-based permutation, the line of work this scheme extends with FAPS."},{"cited_title":"Srss: A new chaos-based single-round single s- box image encryption scheme for highly auto-correlated data,","cited_arxiv_id":null,"evidence_quote":"Provides a prior chaos-based single-round S-box encryption baseline aimed at highly auto-correlated image data."},{"cited_title":"Chaos based image encryption scheme to secure sensitive multimedia content in cloud storage,","cited_arxiv_id":null,"evidence_quote":"Represents an existing chaos-based image encryption approach for cloud storage that motivates the proposed lightweight design."},{"cited_title":"A novel cosine-modulated-polynomial chaotic map to strengthen image encryption algorithms in iot environments,","cited_arxiv_id":null,"evidence_quote":"Shows a chaos-based encryption strengthening technique for IoT environments, supporting the use of chaotic maps in this setting."},{"cited_title":"On the features and challenges of security and privacy in distributed internet of things,","cited_arxiv_id":null,"evidence_quote":"Establishes the security and privacy challenges in distributed IoT networks that the paper uses to motivate the scheme."}],"review_version":1}