Private key and password protection by steganographic image encryption

Debesh Choudhury; Sujoy Chakraborty

REVIEW 4 major objections 5 minor 14 references

Private key and password protection by steganographic image encryption

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that a private key or passcode can be preserved inside an ordinary color photograph and recovered exactly, without the user having to memorize or write down the key.

desk verdict The demo works, but the security claim collapses on LFSR linearity and unanalyzed seed entropy; worth a reviewer's time only to put the limitations on record. read the letter →

arxiv 2507.21068 v1 pith:NODN3U32 submitted 2025-06-05 cs.CR

Debesh Choudhury , Sujoy Chakraborty This is my paper

classification cs.CR

keywords privatekeyprotectionsteganographyQRcodeLFSRimageencryptionvisualpasswordrecoverycryptography

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a private key or passcode can be preserved inside an ordinary color photograph and recovered exactly, without the user having to memorize or write down the key. The plaintext is first turned into a QR code, then encrypted with a stream cipher whose key is generated from a user-chosen visual password image, and finally concealed inside a cover image by steganography. The user recovers the key by remembering the seed image, which the paper argues is easy because people rarely forget images from their own episodic life. If the scheme holds up, it offers a practical answer to the private key-loss conundrum: instead of writing down or memorizing a key, the user keeps a memorable photo and regenerates the key from it. The reported experiment shows a full encode-and-decode round trip on a sample private key.

What carries the argument

The mechanism that carries the scheme is a 128-bit Linear Feedback Shift Register (LFSR), a shift register whose output is a deterministic linear bit sequence. Its initial state is fixed by the user's seed image: each color channel is thresholded at intensity 128, bits are read row-wise in alternating red, green, blue order, and the first 128 bits become the register state. That state generates a keystream which is XORed with the column-wise bits of the QR code, encrypting it into noise; the same seed image regenerates the identical keystream during recovery. A separate steganographic step hides the encrypted image inside the blue channel of a cover photo, and the QR decode step turns the recovered plaintext back into the original key or passcode. The LFSR's maximal-length property, giving $2^{128}-1$ output bits, is what the paper leans on for its brute-force security claim.

What would settle it

Gather a large set of ordinary photographs, threshold their red, green, and blue channels at 128, read the first 128 bits in the paper's alternating-channel order, and count how many distinct initial states occur; if the number is far below $2^{128}$, or if the states are predictable from the image's public content, the brute-force claim fails. A sharper test is to take any public photo as a candidate seed image, obtain the keystream from a known QR-code/ciphertext pair, and show that the same keystream decrypts another image encrypted under the same seed.

Watch

Extended reading notes

Core claim

The paper's central claim is that a private key can be hidden in a real-life color image and later recovered exactly, with no need to remember a passphrase. The chain is: plaintext key to QR code, QR code to white-noise ciphertext by XOR with the output of a 128-bit Linear Feedback Shift Register (LFSR) initialized from thresholded RGB channels of the user's seed image, and ciphertext to a stego image by concealing it in the blue channel of a cover photo. Decryption runs the chain backward using the same seed image. The paper further asserts that a 128-bit LFSR with maximum sequence length makes brute-force attack impossible, and that the innocuous look of the stego image, combined with two-stage cryptographic processing, makes the hidden key extremely difficult to guess or break. These assertions, not the round-trip demonstration alone, are what the paper is trying to establish.

Load-bearing premise

The load-bearing premise is that the first 128 bits obtained by thresholding a user-selected photo at intensity 128 are unpredictable and secret enough to act as a cryptographic key; the paper provides no entropy, correlation, or predictability analysis of those bits.

Editorial extensions

If this is right

Users could back up a private key by storing only ordinary photographs; the key regenerates from the seed image instead of being memorized or written down.
Anyone who extracts the hidden image from the cover photo obtains white-noise ciphertext, not the key, because the QR code is encrypted before concealment.
Recovery requires only the stego image and the same seed image, so keeping multiple copies of the seed image would let a user restore it even if an attacker alters one copy.
The same pipeline applies to any long or complex text passcode, not only cryptographic private keys.
The scheme can be made stronger by swapping the single LFSR for a multi-LFSR stream cipher such as Trivium or A5/1, as the paper itself notes.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The effective security rests on the entropy and secrecy of the 128 thresholded bits taken from the seed image; the paper measures neither, so the true key space could be far smaller than $2^{128}$ if ordinary photos produce correlated bit streams.
Because the LFSR is a linear stream cipher, a known-plaintext situation that reveals the keystream for one QR code would let an attacker decrypt any other payload encrypted with the same seed image; a nonlinear or keyed cipher would close that gap.
An attacker who can guess a shortlist of a user's memorable life photos could threshold each candidate and try to decrypt the stego image, so the practical security depends on the obscurity of the seed image, not only on the register length.
The paper leaves implicit that the derived key depends on exact pixel values and image dimensions, so any resizing, re-encoding, or editing of the seed image would change the key; practical use would need some error tolerance or redundant key material.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The pipeline is new for a backup tool, and the paper is honestly written: conversion of the private key to a QR code, XOR with an LFSR keystream seeded by a thresholded image, and stego concealment in a color photo. The roundtrip works on one sample, and the idea of a memorable image as a key-ingredient carrier is genuinely user-friendly. I would not dismiss the concept out of hand.

The soft spots are not minor. The central security claim in Section 2.3, that a 128-bit LFSR makes brute force impossible, is wrong in a way any cryptography undergraduate should catch. An LFSR is linear over GF(2); 128 bits of known plaintext at known positions recover the initial state by solving linear equations. QR codes leak exactly this: fixed finder patterns at known corners. The stego layer is the only barrier between the attacker and those ciphertext bits, and the paper does not analyze stego security against detection or extraction. The seed image entropy is also never analyzed—thresholded RGB bits from a personal photo are likely highly correlated and guessable, and the paper calls the seed a "visual password" without studying the actual keyspace. The authors' own admission that a single LFSR could be replaced by Trivium or A5/1 concedes the point.

The paper does what it claims as a proof of concept, which is credit. But the claim that this "permanently solves" the private key-loss conundrum is not supported, and the security argument is false as stated. The stego algorithm is a black-box citation; no imperceptibility or capacity analysis is given. One successful roundtrip on a single key is a feasibility note, not a security evaluation.

Who gets value from this? Readers interested in novel backup schemes or in how not to argue stream-cipher security. It deserves a serious referee only because a competent reviewer can articulate the known-plaintext attack and force a revision that either replaces the LFSR with a proper keystream generator and analyzes the seed keying material, or reframes the paper as a usability proposal with honest limitations. As written, reject; the idea might survive a rewrite.

Referee Report

4 major / 5 minor

Summary. The paper proposes a scheme for protecting a private key or passcode by converting it into a QR code, encrypting the QR code with a stream cipher based on a 128-bit LFSR whose initial state is derived from thresholded color channels of a user-selected seed image, and then hiding the encrypted QR code inside a cover color image using a cited steganographic algorithm. Recovery consists of de-concealing the encrypted QR code, decrypting it with the regenerated LFSR keystream, and reading the QR code. The authors report a single successful roundtrip experiment and claim that the 128-bit LFSR makes brute-force attack impossible and that the technique can solve the private key-loss conundrum.

Significance. If the security claims were correct, the scheme would offer a convenient way to recover private keys from memorable images, which could be practically useful for cryptocurrency and password management. However, the central security argument is not substantiated and is in fact invalid for the proposed cipher. The paper contains no formal analysis, no entropy evaluation of the key schedule, no attack model, and only one feasibility experiment. The authors do credit the cited steganography and QR tools, and the encoding/decoding pipeline is clearly described and easy to reproduce, but the core cryptographic guarantee is unsupported.

major comments (4)

[§2.2.3–§2.2.5] The claim that a 128-bit LFSR 'ensures the security of this encryption scheme' because its period is 2^128−1 is invalid. An LFSR is a linear system over GF(2): if an adversary obtains 128 consecutive known plaintext bits (or any 128 bits with known positions), the initial state can be recovered by solving linear equations, regardless of how the state was generated. QR codes provide exactly this known plaintext because every QR code contains fixed finder patterns at known locations, and §2.2.6 fixes the scanning order as column-wise. An adversary who obtains the encrypted QR image—for example by reversing the cited steganography or because the stego image is compromised—can therefore recover the keystream and decrypt the QR code. This flaw is independent of the entropy of the seed image and undermines the paper's main security claim.
[§2.2.3–§2.2.5] The derivation of the 128-bit LFSR initial state from thresholded color channels of a natural image is not analyzed for entropy, correlation, or predictability. Natural images are highly structured, and thresholding at intensity 128 followed by interleaving the first bits of each channel does not guarantee 128 bits of unpredictable key material. The paper gives no entropy estimate, no statistical tests, and no discussion of how the seed image should be chosen to avoid low-entropy or guessable states. Since the seed image is user-selected from episodic memories, an attacker with knowledge of the user's likely images may be able to narrow the key space dramatically.
[§3] The experimental validation consists of a single private key and a single cover image. There is no evaluation of stego capacity, no robustness test against JPEG compression or other image processing, no statistical comparison between the cover and stego images, and no attempt to launch any of the attacks discussed in the paper (e.g., brute force, known-plaintext, or steganalysis). The conclusion that the technique is 'feasible' is therefore supported only as a lossless roundtrip demonstration, not as a security or usability result.
[§3 (future work note)] The authors' own note that the single LFSR 'could be made further stronger to prevent possible attacks on the LFSR, by replacing the single LFSR with combinations of multiple LFSR-s (such as the Trivium or the A5/1 cipher)' concedes that the current cipher is not adequate as a keystream generator. This admission directly contradicts the unconditional security assertion in §2.3 and indicates that the load-bearing security claim cannot be repaired by local parameter changes.

minor comments (5)

[§1] There is a typo: 'or loosing' should be 'or losing'.
[§2.3] The expressions '2 m − 1' and '2 128 − 1' appear without superscript formatting; they should read '2^m − 1' and '2^128 − 1'.
[§2.2.6] The claim that 'The output of LFSR has excellent stochastic property of randomness' is misleading: an LFSR's output is linear and not cryptographically secure, as the known-plaintext attack in the major comments shows.
[§2.3] The steganographic algorithm from reference [6] is cited but not described; the paper should at least specify the embedding capacity, the modification strategy in the blue channel, and the expected distortion, since these affect whether the encrypted QR code can be reliably extracted.
[§4] The conclusion 'This technique can permanently solve the chronic problem of private key-loss conundrum' is overbroad; the reported experiment covers one roundtrip, not a deployment-scale validation, and the security concerns above preclude such a strong claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the encryption/decryption chain is self-contained; apparent security weaknesses are correctness concerns, not circular reductions.

full rationale

The paper does not fit parameters to data, does not generate predictions from fitted constants, and does not define its target result into its inputs. The LFSR initial state is a hand-picked design choice obtained by thresholding the seed image at intensity 128 and taking the first 128 bits; the recovered private key is not used to define that key, and the QR-code encryption/decryption steps are explicit and invertible. The security claim equating the 2^128−1 LFSR period with brute-force impossibility is a cryptographic correctness argument, not a circular one, and the same holds for the known-plaintext concern on QR finder patterns. The only author self-citation, reference [3], merely names the 'private key-loss conundrum' and is not load-bearing for the technical derivation. The Section 3 note that the single LFSR could be replaced by Trivium or A5/1 is a stated limitation rather than a circular step. No equation or claim in the paper reduces by construction to its own inputs.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the unexamined strength of the seed-image-derived key and on the cited stego algorithm, both treated as black boxes.

free parameters (2)

threshold_value_128 = 128
Pixels in each color channel are binarized at intensity 128; this design constant is chosen by the authors, not derived from the data (Section 2.2.3).
lfsr_length_128 = 128 bits
The LFSR is chosen to be 128 bits; the paper claims brute-force infeasibility based on this length but does not analyze the actual key entropy (Section 2.2.5).

assumptions (3)

domain assumption The referenced steganographic algorithm [6] can hide the encrypted QR code in the blue channel and recover it exactly.
The paper cites Kini and Kini for the stego method without describing its capacity, robustness, or embedding parameters (Section 2.3).
domain assumption The first 128 bits from the thresholded seed image are reproducible each time the user presents the same image.
The decoder regenerates the LFSR state from the seed image; this requires pixel-identical reacquisition or exact restoration, which is assumed in Section 2.2.5 and Section 1.
standard math A 128-bit LFSR with maximum sequence length has a period of 2^128 - 1, making brute force infeasible.
Standard property of maximal-length LFSRs, cited from [5] and used in Section 2.3; it does not account for partial state recovery via known-plaintext attacks.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Private key and password protection by steganographic image encryption." pith.science (2026). https://pith.science/paper/NODN3U32

@misc{pith2026250721068,
  author       = {Pith},
  title        = {Pith review of: Private key and password protection by steganographic image encryption},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NODN3U32}},
  note         = {Machine review of arXiv:2507.21068}
}

read the original abstract

We propose a technique to protect and preserve a private key or a passcode in an encrypted two-dimensional graphical image. The plaintext private key or the passcode is converted into an encrypted QR code and embedded into a real-life color image with a steganographic scheme. The private key or the passcode is recovered from the stego color image by first extracting the encrypted QR code from the color image, followed by decryption of the QR code. The cryptographic key for encryption of the QR code is generated from the output of a Linear Feedback Shift Register (LFSR), initialized by a seed image chosen by the user. The user can store the seed image securely, without the knowledge of an attacker. Even if an active attacker modifies the seed image (without knowledge of the fact that it is the seed image), the user can easily restore it if he/she keeps multiple copies of it, so that the encryption key can be regenerated easily. Our experiments prove the feasibility of the technique using sample private key data and real-life color images.

Figures

Figures reproduced from arXiv: 2507.21068 by the authors.

**Figure 2.** A General LFSR of m bits. Adopted from [5]. 2.2.5 Initialize the LFSR The first 128 bits of the generated bit stream is then used to set the initial state of a 128-bit LFSR. The output of the LFSR is then used to encrypt the QR code, which is also a binary image. 2.2.6 Encrypting the QR code The QR code is encrypted with the 128-bit binary output of the LFSR.5 Let’s say the QR code is of size M × N. First, the QR co… view at source ↗

**Figure 3.** Experimental results: (a) Sample private key, (b) QR Code of the private key, (c) Encrypted QR Code, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

14 extracted references · 12 canonical work pages

[1]

A Method for Obtaining Digital Signatures and Public-Key Cryp- tosystems

R. Rivest, A. Shamir and L. Adleman, “A Method for Obtaining Digital Signatures and Public-Key Cryp- tosystems”, Communications of the ACM. 21 (2): 120–126 (February 1978)

work page 1978
[2]

Stinson and M

D. Stinson and M. Paterson, Cryptography: Theory and Practice, 4th Edition (CRC Press, August 2018)

work page 2018
[3]

Private Key-Loss Conundrum: The roadblock to Blockchain’s acceptance by the mass

D. Choudhury,“Private Key-Loss Conundrum: The roadblock to Blockchain’s acceptance by the mass”, https://www.linkedin.com/pulse/private-key-loss-conundrum-blockchain-roadblock-mass-debesh/ (Accessed July 8, 2022)

work page 2022
[4]

Episodic memory: from mind to brain

E. Tulving, “Episodic memory: from mind to brain”, Annu Rev Psychol. 2002;53:1-25. doi: 10.1146/an- nurev.psych.53.100901.135114. PMID: 11752477

arXiv 2002
[5]

A note on cyclotomic polynomials and Linear Feedback Shift Registers

L. Capuano and A. J. Di Scala, “A note on cyclotomic polynomials and Linear Feedback Shift Registers”, Quaestiones Mathematicae (September 2021). pp.1-13. https://doi.org/10.2989/16073606.2021.1967504

arXiv 2021
[6]

A Secured Steganography Algorithm for Hiding an Image in an Image

N.G. Kini and V.G. Kini, “A Secured Steganography Algorithm for Hiding an Image in an Image”, Integrated Intelligent Computing, Communication and Security, pp. 539–546. Springer, Singapore (2019)

work page 2019
[7]

Creating and Reading QR Codes

T. Sch¨ urmann, “Creating and Reading QR Codes”, Linux Magazine https://www.linux- magazine.com/Online/Features/Generating-QR-Codes-in-Linux (Accessed July 8, 2022)

work page 2022
[8]

Paar and J

C. Paar and J. Pelzl, Understanding Cryptography, (Springer 2009)

work page 2009

Show all 14 references

[9]

A modified stream generator for the GSM encryption algorithms A5/1 and A5/2

I. Erguler and E. Anarim, “A modified stream generator for the GSM encryption algorithms A5/1 and A5/2”, 13th European Signal Processing Conference (2005)

2005
[10]

An example private key, https://en.bitcoin.it/wiki/Private key
[11]

Bitcoin: A Peer-to-Peer Electronic Cash System

S. Nakamoto, “Bitcoin: A Peer-to-Peer Electronic Cash System”, https://bitcoin.org/bitcoin.pdf (Accessed July 8, 2022)

2022
[12]

A. M. Antonopoulos, Mastering Bitcoin: Unlocking Digital Cryptocurrencies (O’Reilly Media, 2015)

2015
[13]

Wattenhofer, Distributed Ledger Technology: The Science of the Blockchain (Createspace Independent Publishing, March 2017)

R. Wattenhofer, Distributed Ledger Technology: The Science of the Blockchain (Createspace Independent Publishing, March 2017)

2017
[14]

McDonald, The Password Book: Internet Security & Passwords Made Easy (CreateSpace Independent Publishing Platform, August 2017)

J. McDonald, The Password Book: Internet Security & Passwords Made Easy (CreateSpace Independent Publishing Platform, August 2017)

2017

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

A Method for Obtaining Digital Signatures and Public-Key Cryp- tosystems

R. Rivest, A. Shamir and L. Adleman, “A Method for Obtaining Digital Signatures and Public-Key Cryp- tosystems”, Communications of the ACM. 21 (2): 120–126 (February 1978)

work page 1978

[2] [2]

Stinson and M

D. Stinson and M. Paterson, Cryptography: Theory and Practice, 4th Edition (CRC Press, August 2018)

work page 2018

[3] [3]

Private Key-Loss Conundrum: The roadblock to Blockchain’s acceptance by the mass

D. Choudhury,“Private Key-Loss Conundrum: The roadblock to Blockchain’s acceptance by the mass”, https://www.linkedin.com/pulse/private-key-loss-conundrum-blockchain-roadblock-mass-debesh/ (Accessed July 8, 2022)

work page 2022

[4] [4]

Episodic memory: from mind to brain

E. Tulving, “Episodic memory: from mind to brain”, Annu Rev Psychol. 2002;53:1-25. doi: 10.1146/an- nurev.psych.53.100901.135114. PMID: 11752477

arXiv 2002

[5] [5]

A note on cyclotomic polynomials and Linear Feedback Shift Registers

L. Capuano and A. J. Di Scala, “A note on cyclotomic polynomials and Linear Feedback Shift Registers”, Quaestiones Mathematicae (September 2021). pp.1-13. https://doi.org/10.2989/16073606.2021.1967504

arXiv 2021

[6] [6]

A Secured Steganography Algorithm for Hiding an Image in an Image

N.G. Kini and V.G. Kini, “A Secured Steganography Algorithm for Hiding an Image in an Image”, Integrated Intelligent Computing, Communication and Security, pp. 539–546. Springer, Singapore (2019)

work page 2019

[7] [7]

Creating and Reading QR Codes

T. Sch¨ urmann, “Creating and Reading QR Codes”, Linux Magazine https://www.linux- magazine.com/Online/Features/Generating-QR-Codes-in-Linux (Accessed July 8, 2022)

work page 2022

[8] [8]

Paar and J

C. Paar and J. Pelzl, Understanding Cryptography, (Springer 2009)

work page 2009