Bio-Inspired Cyber Defense: Enhancing Intrusion Detection

through the Waterwheel Plant Optimization Algorithm

Omar Bani Fayyad 1,*

1 School of ICT, Bahrain Polytechnic, PO Box 33349, Isa Town, Bahrain

Received: May 16, 2026 Revised: June 25, 2026 Accepted: August 27, 2026 ⋆ Corresponding author

ABSTRACT

In today’s increasingly interconnected digital landscape, intrusion detection systems (IDS) are critical for ensuring

the security and integrity of network infrastructures. However, the effectiveness of traditional IDS models can be

hampered by challenges such as high-dimensional data and the complex dynamics of attack patterns. This paper

addresses these challenges by leveraging the Waterwheel Plant Optimization Algorithm (WWPA), a bio-inspired

metaheuristic, to optimize the hyperparameters of the Support Vector Machine (SVM), a widely used classifier for

intrusion detection. Our contributions include the development of an optimized machine learning framework that

significantly improves classification performance by integrating WWPA with SVM. Experimental results show that

the baseline SVM achieves an accuracy of 86.36%, while the WWPA-optimized SVM yields an impressive accuracy

of 97.00%. These results highlight the efficacy of WWPA in navigating the complex hyperparameter search space

and enhancing the sensitivity and precision of intrusion detection models. The implications of this work are profound

for cybersecurity, as it provides a robust, efficient, and adaptive solution for detecting intrusions with high accuracy,

thereby improving the reliability and responsiveness of modern IDS systems.

Keywords: Intrusion Detection Systems Waterwheel Plant Algorithm (WWPA) Support Vector Machine (SVM)

Optimization Metaheuristic Optimization Bio-Inspired Algorithms

1. INTRODUCTION

Today, the scope of penetration is growing, but this is happening

because of the interconnected systems, clouds, and

platforms that provide lots of data and are the target for the

confidentiality, integrity, and availability of information.[1]

Technological development elevated cybersecurity and network

defense systems to become vital necessities for all modern

technological frameworks.[2] Powered by the enormous

and intricate network surface area and the energetic murkiness

of cyber dangers, programmed and insightful frameworks that

can spot, respond to, and, if possible, stop malicious or bots

from getting underway naturally are essential[3].

Intrusion detection systems (IDS) are fundamental components

that cannot be disregarded in this situation.[3] These are

sentinels of digital infrastructures that aim to detect involuntary

access or misuse on networks and anomalies of behavior.

Proactive and effective countermeasures can be applied, thus

allowing risk mitigation and the system’s functionality with

trust. Especially, any detection in the early stage of the intrusion

limits the attack’s spread and damage, system downtime,

and prevents data exfiltration or corruption.[4] Cyber attacks

are becoming more complex and more frequent, and more

sophisticated, so the traditional rule-based detection system

has no idea how to come up with a solution to very novel,

obfuscated attack strategies anymore[5]. There’s a need for

something adaptive, intelligent.

Machine learning (ML) established itself as a disruptive