Hybrid Firefly–Particle Swarm Optimization for Parameter Tuning of XGBoost: Network Intrusion Detection

Abstract Machine learning-based Intrusion Detection Systems (IDS) have demonstrated strong detection performance but remain highly sensitive to hyperparameter choices. XGBoost is widely used for tabular security data but requires careful tuning to avoid overfitting and to minimize false positives. This paper presents a practically oriented hybrid Firef ly–Particle Swarm Optimization (FA–PSO) framework to tune critical XGBoost hyperparameters for binary network intrusion detection on NSL-KDD. Weexpand the state-of-the-art review adopting recent works (2022–2025), provide full implementation details (hardware/software, parameter ranges, reproducible seeds), present algorithm pseudo-code, and evaluate performance against baseline tuning methods. Results indicate the hybrid method converges faster and yields improved detection metrics while reducing analyst workload via lower false positive rates. The manuscript includes code-level descriptions enabling reviewers to reproduce experiments and facilitating journal-level publication standards