article · Journal africain des sciences.
This study presents a comparative evaluation of three deep learning architectures (RNN, CNN, and MLP) for early intrusion detection using the NSL-KDD and UNSW-NB15 datasets. The results show that, on NSL-KDD, the RNN achieves the best overall performance (accuracy ≈ 77.9%; F1-score ≈ 77.5%), while the CNN stands out with high sensitivity (≈ 94.2%), despite a moderate recall indicating the presence of false negatives. In contrast, on UNSW-NB15, all models achieve high overall performance (accuracy > 93%). The RNN (LSTM) demonstrates strong effectiveness with high sensitivity (≈ 98.8%) and F1-score (≈ 94.9%), confirming its suitability for proactive intrusion detection. The CNN attains the highest precision (≈ 95.8%), whereas the MLP shows a high recall (≈ 95.3%). Overall, these findings confirm the effectiveness of Deep Learning approaches for intrusion detection, with a notable advantage of recurrent models in capturing the temporal dynamics of network traffic.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.70237/jafrisci.2026.v3.i4.05
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.