Detecting Hidden Data in Images Using Convolutional Neural Networks
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Recent advancements in Deep Learning (DL) have driven the development of innovative methodologies, particularly within the domain of steganalysis for spatial domain images.Steganalysis, as the counterpart to steganography, is dedicated to uncovering concealed data within the content, making a digital image.Convolutional Neural Networks (CNNs), grounded in DL principles, have been influential in pushing the boundaries of this field.Despite the development of various CNN architectures that have raised the precision in detecting images with steganographic payload, current models contend with challenges related to the detectability of low payload capacities and suboptimal processes for feature learning.In response, this study introduces a novel CNN architecture to enhance steganalysis and improve the accuracy of detecting covert data in spatial domain images.The proposed model introduces a strategic integration of maximum and average pooling, a tandem approach meticulously designed to amplify the network's proficiency in capturing intricate details and multiple layers of information.Moreover, the proposed CNN architecture is structured into three principal stages: preprocessing, feature extraction, and classification.The preprocessing stage comprises Input, regular convolution layer, and Batch Normalization.The feature extraction stage employs the ReLU as a non-linear activation function based on its capacity to expedite computation by bypassing the need for exponentials and divisions.The classification stage introduces the multi-scale inception module to enhance the probabilistic feature classification.The proposed model's correctness in probabilistic classification through the receiver operating characteristic curve (ROC AUC) yields an AUC of 0.95, reflecting a prediction correctness of 95%.Furthermore, the results show that the proposed model outperforms the results of previous research studies in terms of accuracy and improves the existing works with a percentage ranging from 2.3 to 2.9%.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it