Lightweight image steganalysis with block-wise pruning.

Image steganalysis is the task of detecting a secret message hidden in an image. Deep steganalysis using end-to-end deep learning has been successful in recent years, but previous studies focused on improving detection performance rather than designing a lightweight model for practical applications. This caused a deep steganalysis model to be heavy and computationally costly, making the model infeasible to deploy in real-world applications. To address this issue, we study an effective model design strategy for lightweight image steganalysis. Considering the domain-specific characteristics of steganalysis, we propose a simple yet effective block removal strategy that progressively removes a sequence of blocks from deep classification networks. This method involves the gradual removal of convolutional neural network blocks, starting from deeper ones. By doing so, the number of parameters and FLOPs are decreased without compromising the detection performance. Experimental results show that our removal strategy makes the EfficientNet-B0 variants 9.58 [Formula: see text] smaller and has 2.16 [Formula: see text] fewer FLOPs than the baseline while retaining detection accuracy of 90.73% and 82.40% that are on par with the baseline on BOSSBase and ALASKA#2 datasets, respectively. Backed by our in-depth analyses, the results indicate that only a few early layers are sufficient for effective image steganalysis.

Predicting effects of built environment on fatal pedestrian accidents at location-specific level: Application of XGBoost and SHAP.

Understanding locally heterogeneous physical contexts in built environment is of great importance in developing preemptive countermeasures to mitigate pedestrian fatality risks. In this study, we aim to investigate the non-linear relationship between physical factors and pedestrian fatality at a location-specific level using a machine learning approach. The state-of-art machine learning algorithm, eXtreme Gradient Boosting (XGBoost), is employed for a binary classification problem, in which nationwide locations where fatal pedestrian accidents occurred for the years from 2012 to 2019 in Korea serve as positive samples (np = 13,366). For negative samples, locations with no pedestrian accidents are selected randomly to the size that is 10 times larger (nn = 133,660) than positive samples. Fifteen features under the categories of road conditions, road facilities, road networks, and land uses are assigned to both the positive and negative sample locations using Geographic Information System (GIS). A method is proposed to avoid the class imbalance problem, and a final unbiased model is utilized to predict fatal pedestrian risks at the negative sample locations. In addition, Shapley Additive Explanations (SHAP) is introduced to provide a robust interpretation of the XGBoos prediction results. It is shown that 21.6% of the negative sample locations have a probability of fatal pedestrian accidents greater than 0.5 (or 78.4% accuracy). Generally, a road segment that lies in many of the shortest routes in a dense residential area with many lively activities from aligned buildings is a potential spot for fatal pedestrian accidents. However, based on the SHAP interpretation, the relationships between the features and pedestrian fatality are found nonlinear and locally heterogeneous. We discuss the implications of this result has for drafting policy recommendations to reduce pedestrian fatalities.