One moment / Route transition
The page is being assembled. Navigation remains available while the content catches up.
Research 03 / Image encryption
A graduation research project combining nonlinear dynamics, a dynamic RGB S-box, and a desktop application to investigate reversible color-image encryption.
Inspect the research
This is an academic proof of concept. The reported experiments are presented as research evidence, not as an independently audited or production-ready cryptographic claim.
B.Sc. graduation research · 2024
Model · algorithm · UI · experiments
SciPy simulation · NumPy transformation
Experimental results · no production claim
Research dossier / Method
01—04
Next: how five coupled nonlinear equations become a dynamic RGB substitution box, then a reversible pixel transformation.
01 / Transformation method
01 / METHOD
Five coupled nonlinear state variables evolve from the supplied initial conditions. Their trajectories feed an RGB substitution box, which becomes the repeatable key field used by the image transform.
The numerical source
5 STATE VARIABLES
The system couples x, y, z, u, and v through linear, multiplicative, and sinusoidal terms. Small changes to the starting state produce a different numerical trajectory for the experiment.
x′ = −10.2x + 12y − sin(u)v
y′ = −5.1y + 30x − 2.5xz − sin(v)x
z′ = −6z + xy + 3sin(x)v
u′ = −u − yz − 5sin(y)x
v′ = −1.1v − 4zu + sin(z)y
Implementation record
The displayed equations and solver settings are transcribed from the working Python implementation.
From state to pixels
Each stage has a specific job: establish the state, sample it, map it into RGB values, transform the image, then perform the inverse in the opposite order.
The user supplies initial values for x, y, z, u, and v together with an S-box size.
SciPy numerically solves the coupled equations and returns five evolving sequences.
Derived x, y, and z sequences are reduced to 8-bit channel values and arranged as a color S-box.
The S-box repeats across the image. Each channel is XORed with its key field, then shifted modulo 256.
Decryption subtracts the value shift first, then applies the same XOR key field to recover the pixels.
Forward / inverse
P is the source-channel value, K is the corresponding tiled S-box value, s is the configured shift, and C is the encrypted value.
Encryption
C = ((P XOR K) + s) mod 256
Decryption
P = ((C − s) mod 256) XOR K
The implementation performs a modular pixel-value shift. It does not spatially rotate or rearrange pixel positions.
Engineering decision / Vectorization
The first approach expressed the algorithm as nested channel, row, and column loops. The working path tiles the S-box with NumPy and transforms complete channel arrays, reducing Python-level work while preserving the same inverse operation.
02 / Experimental evidence
02 / REPORTED RESULTS
The research evaluated statistical behavior, sensitivity, visual similarity, recovery, and execution time. This chapter keeps the reported results visible alongside the limits of their interpretation.
Reported result register
08 REPORTED SIGNALS
No single score validates an encryption design. Read together, the measurements describe the output distribution, response to change, recovery fidelity, and observed runtime.
Reported information density near the 8-bit upper bound.
Reported percentage of changed encrypted pixels under the paper's comparison.
Reported average intensity change between compared cipher images.
Reported adjacent-pixel coefficient in the encrypted image.
Reported adjacent-pixel coefficient in the encrypted image.
Low reported similarity between the source and encrypted image.
Original-to-decrypted structural similarity reported as exact recovery.
Typical reported runtime for the paper's representative experiment.
Values are reported to their paper precision. Precision in the table does not imply independent verification or broader generalization.
Interpretation discipline
The useful story is not that every value is high or low. It is that each metric answers a narrower question about the experiment.
What it supports
Entropy near 8 bits, near-zero adjacent correlation, and low original-to-encrypted PSNR are consistent with a visually flattened output unlike its source.
What it does not prove
Statistical appearance alone does not establish resistance to cryptanalysis, key recovery, or structural attacks.
What it supports
The reported NPCR and UACI values indicate that the tested comparison produced broad pixel and intensity changes.
What it does not prove
A representative differential test is not a substitute for a defined threat model or systematic adversarial evaluation.
What it supports
An original-to-decrypted SSIM of 1.0 supports lossless recovery for the reported test image and parameters.
What it does not prove
Successful inversion confirms implementation behavior, not confidentiality or authenticity under attack.
What it supports
The reported 0.7–0.8 second encryption time shows the vectorized prototype was interactive in its test setting.
What it does not prove
It is not a cross-device benchmark, scale guarantee, or real-time performance claim.
Artifact / Working prototype
Original, encrypted, generated S-box, and decrypted states appear together, allowing the researcher to run and visually review the complete transformation loop.
This artifact demonstrates a working end-to-end desktop experiment; it does not independently validate cryptographic security.
Research limits
The supplied results describe representative 256 × 256 color-image experiments, not a broad standardized corpus.
This portfolio records the paper's measurements. It does not present a fresh independent benchmark run or reproducibility package.
The system remains an academic proof of concept and should not be treated as suitable for security-critical production use.
03 / Productization
03 / DESKTOP INSTRUMENT
The desktop application connects configuration, transformation, visual inspection, measurement, and export in one controlled workflow. The interface is not decoration around the algorithm; it is how the experiment becomes repeatable and understandable.
Interface model
The main workspace places source, cipher, generated S-box, and recovered image side by side. Inputs and actions stay beneath the evidence, while a separate Evaluation tab handles measurement and plotting.
Original image
01Encrypted image
02RGB S-box
03Decrypted image
04Interface reconstruction based on the working application structure; it explains information architecture rather than replacing the real screenshot above.
The researcher workflow
The product follows the sequence of an experiment instead of exposing implementation modules as disconnected controls.
Load an image, normalize it to RGB when needed, enter five initial conditions, and choose an S-box size.
Encrypt and decrypt on worker threads while progress feedback keeps the desktop interface responsive.
Compare four image states, update metrics, view three histograms, and plot chaotic-system trajectories.
Save encrypted, decrypted, and S-box PNG files—or reload an encrypted image with matching parameters for decryption.
Capability map
C-01
C-02
C-03
C-04
Software layers
A modular structure lets the GUI coordinate work without owning the chaotic model, transformation logic, or evaluation formulas.
Tabs, inputs, previews, progress, dialogs, and theme switching.
Image state, S-box generation, forward and inverse transforms, and export methods.
Vectorized channel operations and numerical integration of the five-dimensional system.
Image I/O, similarity and statistical evaluation, histograms, and trajectory plots.
Windows executable packaging with runtime-safe paths for bundled resources.
Product engineering decisions
Keeping all four image states visible makes reversibility inspectable and reduces the mental work of comparing separate files or windows.
Encryption and decryption run in worker threads; UI updates return through Tkinter's event loop so progress remains visible during computation.
The interface checks missing images, numeric initial conditions, positive S-box size, and image-size bounds before starting the transformation.
04 / Research outcome
04 / OUTCOME RECORD
Delivered outcome
Graduation research+ working Windows prototype
The project joins a proposed five-dimensional chaotic model, RGB S-box generation, reversible image transformation, experimental evaluation, desktop interaction, technical writing, and executable packaging in one authored body of work.
Verified against the supplied brief, working Python source, application artifact, and owner-provided research record.
Completed as research software; not independently audited or approved for security-critical production use.
End-to-end contribution
MODEL → WINDOWS DELIVERY
The strongest signal is not one metric. It is the ability to move across research, numerical implementation, image processing, evaluation, interface design, and delivery without losing the thread between them.
Develop the five-dimensional chaotic model and define how its state can feed an image-encryption experiment.
Build the RGB S-box, reversible transform, vectorized image path, and modular Python structure.
Implement statistical and image-quality measures, conduct experiments, interpret results, and document the research.
Design the desktop workflow, add inspection and export tools, and package the application for Windows.
What serious continuation requires
RESEARCH ROADMAP / NOT SHIPPED
Progress from prototype to credible security research would require a reproducible evaluation system, broader comparisons, explicit threat modeling, and scrutiny from specialists outside the project.
Create deterministic experiment scripts, fixtures, parameter records, and regression tests so every reported result can be rerun.
Evaluate more image sizes, formats, datasets, hardware profiles, competing methods, and adverse parameter choices.
Define the threat model and submit the construction, key space, attack surface, and evaluation method to independent specialists.
Explore acceleration or additional media such as video and audio only after the core design and evaluation process become dependable.
End of research dossier / Start of a conversation
This project represents the kind of engineering I value: translating an unfamiliar technical idea into software people can inspect, question, and use—without hiding the limits of the evidence.