stegosphere.analysis package

Submodules

stegosphere.analysis.accuracy module

Accuracy metrics for comparing original and extracted binary payloads.

stegosphere.analysis.accuracy.bit_error_rate(before, after)

Compute the bit error rate between two binary sequences.

Parameters:

  • before (str | list | numpy.ndarray) – Original binary sequence.

  • after (str | list | numpy.ndarray) – Modified binary sequence.

Returns:

Proportion of bits that differ.

Return type:

float

Raises:

ValueError – If inputs are not the same length.

stegosphere.analysis.accuracy.hamming_distance(before, after)

Compute the Hamming distance between two binary strings or arrays.

Parameters:

  • before (str | list | numpy.ndarray) – Original binary sequence.

  • after (str | list | numpy.ndarray) – Modified binary sequence.

Returns:

Number of differing bits.

Return type:

int

Raises:

ValueError – If inputs are not the same length.

stegosphere.analysis.accuracy.normalized_correlation(before, after)

Compute the normalized correlation between two sequences.

Parameters:

  • before – Original numerical sequence.

  • after – Modified numerical sequence.

Returns:

Normalized correlation coefficient. Returns 0 if denominator is zero.

Return type:

float

Raises:

ValueError – If inputs are not the same length.

stegosphere.analysis.capacity module

Measuring how much data can be stored within a cover.

In future versions, this should include heuristics using the max_capacity_estimator

stegosphere.analysis.capacity.bits_per_value(array, max_capacity)

Compute bits per host‐value given total capacity.

Parameters:

  • array (numpy.ndarray) – Host array.

  • max_capacity (int | float) – Total embedding capacity in bits.

Returns:

Bits of payload per array element.

Return type:

float

stegosphere.analysis.capacity.bits_per_value_used(array, payload_size)

Compute actually used bits per host‐value.

Parameters:

  • array (numpy.ndarray) – Host array.

  • payload_size (int | float) – Size of embedded payload in bits.

Returns:

Payload bits per array element.

Return type:

float

stegosphere.analysis.capacity.max_capacity_estimator(*args)
stegosphere.analysis.capacity.payload_ratio(payload_size, max_capacity)

Compute payload-to-capacity ratio.

Parameters:

  • payload_size (int | float) – Size of embedded payload in bits.

  • max_capacity (int | float) – Maximum capacity in bits.

Returns:

Ratio payload_size / max_capacity.

Return type:

float

stegosphere.analysis.detectability module

A module to detect the likelihood of steganography

stegosphere.analysis.detectability.random_detector(input_data)

Randomly decide presence of steganography (for testing).

Parameters:

input_data – Any input; ignored in decision.

Returns:

0 or 1, chosen uniformly at random.

Return type:

int

stegosphere.analysis.detectability.sp(X)

Estimate embedding rate (β̂) via Sample Pairs Analysis (Dumitrescu et al. 2002).

Parameters:

X (array-like) – 2D array of integer pixel values.

Returns:

Estimated embedding rate β̂ ∈ [0,1], or -1 if undetermined.

Return type:

float

Raises:

ValueError – If input is not a 2D array with at least 2 columns.

stegosphere.analysis.detectability.uniformity_detector(array)

Detect likelihood of LSB steganography in an RGB image by uniformity of LSBs.

Parameters:

array (array-like) – H×W×3 RGB image array of integers.

Returns:

Score ∈ [0,1], where higher means more uniform LSB distribution

(i.e. more likely stego).

Return type:

float

Raises:

ValueError – If input is not a 3D array with third dimension size 3.

stegosphere.analysis.efficiency module

Module for measuring elapsed and used memory of processes.

class stegosphere.analysis.efficiency.EfficiencyProfiler

Bases: object

class stegosphere.analysis.efficiency.MemoryProfiler

Bases: object

class stegosphere.analysis.efficiency.Timer

Bases: object

stegosphere.analysis.efficiency.profile_function(func)

Decorator to time and measure memory of a function call. :returns: result, profiling_info (dict)

stegosphere.analysis.imperceptibility module

Measuring imperceptibility

stegosphere.analysis.imperceptibility.bhattacharyya_divergence(array_1, array_2, epsilon=1e-10)

Compute Bhattacharyya distance between two distributions.

Parameters:

  • array_1 (array-like) – Probability distributions or counts.

  • array_2 (array-like) – Probability distributions or counts.

  • epsilon (float) – Small constant to avoid log(0).

Returns:

Bhattacharyya distance.

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.chi_distance(array_1, array_2, epsilon=1e-10)

Compute chi-squared distance between two nonnegative arrays.

Parameters:

  • array_1 (array-like) – Input histograms or nonnegative arrays.

  • array_2 (array-like) – Input histograms or nonnegative arrays.

  • epsilon (float) – Small constant to avoid division by zero.

Returns:

Chi-squared distance.

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.euclidean_distance(array_1, array_2)

Compute Euclidean (L2) distance between two arrays.

Parameters:

  • array_1 (array-like) – Input arrays of equal shape.

  • array_2 (array-like) – Input arrays of equal shape.

Returns:

Euclidean norm of differences.

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.js_divergence(array_1, array_2, epsilon=1e-10)

Compute Jensen-Shannon divergence between two distributions.

Parameters:

  • array_1 (array-like) – Probability distributions or counts.

  • array_2 (array-like) – Probability distributions or counts.

  • epsilon (float) – Small constant to avoid log(0).

Returns:

JS divergence.

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.kl_divergence(array_1, array_2, epsilon=1e-10)

Compute Kullback-Leibler divergence Dₖₗ(P‖Q).

Parameters:

  • array_1 (array-like) – Probability distributions or counts.

  • array_2 (array-like) – Probability distributions or counts.

  • epsilon (float) – Small constant to avoid log(0).

Returns:

KL divergence.

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.manhattan_distance(array_1, array_2)

Compute Manhattan (L1) distance between two arrays.

Parameters:

  • array_1 (array-like) – Input arrays of equal shape.

  • array_2 (array-like) – Input arrays of equal shape.

Returns:

Sum of absolute differences.

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.mse(array_1, array_2)

Compute Mean Squared Error (MSE) between two arrays.

Parameters:

  • array_1 (array-like) – Input arrays of equal shape.

  • array_2 (array-like) – Input arrays of equal shape.

Returns:

MSE value.

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.psnr(array_1, array_2, max_i=None)

Compute Peak Signal-to-Noise Ratio (PSNR) in decibels.

Parameters:

  • array_1 (array-like) – Input arrays of equal shape.

  • array_2 (array-like) – Input arrays of equal shape.

  • max_i (number, optional) – Maximum possible pixel value. Defaults to dtype max of array_1.

Returns:

PSNR value (inf if zero error).

Return type:

float

Raises:

ValueError – If shapes differ.

stegosphere.analysis.imperceptibility.results(array_1, array_2, measures='all', epsilon=1e-10, psnr_max_i=None)

Compute multiple similarity/divergence/distance measures.

Parameters:

  • array_1 (array-like) – Input arrays of equal shape.

  • array_2 (array-like) – Input arrays of equal shape.

  • measures (list or 'all') – Names of measures to compute.

  • epsilon (float, optional) – Small constant for stability.

  • psnr_max_i (number, optional) – Max value for PSNR.

Returns:

{measure_name: value, …}

Return type:

dict

stegosphere.analysis.imperceptibility.ssim(array_1, array_2, **kwargs)

Compute Structural Similarity Index (SSIM) between two images.

Parameters:

  • array_1 (ndarray) – Grayscale or multi-channel images.

  • array_2 (ndarray) – Grayscale or multi-channel images.

  • **kwargs – Passed to skimage.metrics.structural_similarity.

Returns:

SSIM score (and image if full=True).

Return type:

float or ndarray

Raises:

ValueError – If shapes differ.

stegosphere.analysis.robustness module

A module for simulating attacks onto arrays.

stegosphere.analysis.robustness.cropping_attack(image, crop_fraction=0.1)

Randomly crop and pad back to original size.

Parameters:

  • image (ndarray) – Input image.

  • crop_fraction (float) – Fractional size to crop on each side.

Returns:

Cropped and resized image.

Return type:

ndarray

stegosphere.analysis.robustness.gaussian_noise_attack(array, sigma=1.0)

Apply Gaussian smoothing as a low-pass filter.

Parameters:

  • array (ndarray) – Input image or signal.

  • sigma (float) – Standard deviation for Gaussian kernel.

Returns:

Smoothed result.

Return type:

ndarray

stegosphere.analysis.robustness.jpeg_compression_attack(image, quality=50)

Simulate JPEG compression artifacts.

Parameters:

  • image (ndarray) – Input image in BGR or grayscale.

  • quality (int) – JPEG quality factor (0-100).

Returns:

Compressed-then-decoded image.

Return type:

ndarray

stegosphere.analysis.robustness.random_bit_attack(array, ratio=0.01)

Randomly flip a given fraction of elements in the array by +1 or -1.

Parameters:

  • array (array-like) – Input numeric array.

  • ratio (float) – Fraction of elements to perturb ∈ [0, 1].

Returns:

Copy of the input with random ±1 perturbations applied.

Return type:

np.ndarray

Raises:

ValueError – If ratio is not between 0 and 1.

stegosphere.analysis.visualizer module

Module to visualize various measures. Currently mostly for images.

stegosphere.analysis.visualizer.channel_correlation_plot(data)

Plot correlation between channels of a multichannel array.

Parameters:

data (ndarray) – Multichannel array, last dim = channels.

stegosphere.analysis.visualizer.difference_heatmap(orig, modified, title='Difference Heatmap', cmap='seismic', vmin=None, vmax=None)

Plot heatmap of differences between two arrays.

Parameters:

  • orig (ndarray) – Arrays of the same shape.

  • modified (ndarray) – Arrays of the same shape.

  • title (str) – Plot title.

  • cmap (str) – Colormap.

  • vmin (float or None) – Value range for color scaling.

  • vmax (float or None) – Value range for color scaling.

stegosphere.analysis.visualizer.histogram_comparison(original, modified, bins=256)

Compare histograms of two arrays.

Parameters:

  • original (ndarray) – Input arrays.

  • modified (ndarray) – Input arrays.

  • bins (int) – Histogram bins.

stegosphere.analysis.visualizer.plot_lsb_frequency(data)

Plot LSB frequency of an array (supports multichannel images, audio, etc.).

Parameters:

data (ndarray) – Input array.

stegosphere.analysis.visualizer.show_side_by_side(a1, a2, titles=('Original', 'Modified'))

Show two arrays (e.g., images, spectrograms) side by side.

Parameters:

  • a1 (ndarray) – Arrays to show.

  • a2 (ndarray) – Arrays to show.

  • titles (tuple) – Subplot titles.

Module contents

The analysis package is a collection of various tools for steganalysis.