- https://doi.org/10.1109/icaitech66481.2025.11387490
Optimizing DDoS Attack Detection in IoT Networks Using a Sparse Autoencoder: Work in Progress
- Nov 20, 2025
- Abdul Rahim +5 more
The escalating threat of Distributed Denial of Service (DDoS) attacks poses a significant challenge to the security of Internet of Things (IoT) networks, where resource-constrained devices are frequently exploited by botnets. This research proposes and validates a hybrid model that combines a Sparse Autoencoder (SAE) for feature extraction and dimensionality reduction with a Random Forest (RF) for classification. Utilizing the CICIoT2023 dataset, the key findings show that the SAE efficiently reduced the feature set from 46 to 8 features. Despite a feature reduction of over $80 \%$, the SAE-RF model achieved perfect performance with an F1Score of $\mathbf{1. 0 0 0 0}$, matching the baseline model that used all $\mathbf{4 6}$ features. More importantly, computational efficiency was significantly improved, evidenced by a $25 \%$ reduction in inference time from 0.0008 ms to 0.0006 ms per sample. These results confirm that the computational load can be drastically reduced without sacrificing detection performance, thus validating the model’s suitability for resource-constrained IoT environments.