October 15, 2025 Thesis Open Access

Mesfin Kiflu

JSON-LD (schema.org) Export

{
  "inLanguage": {
    "alternateName": "eng", 
    "@type": "Language", 
    "name": "English"
  }, 
  "description": "<p><strong>Abstract</strong>: Mobile Ad Hoc Networks (MANETs) are a prominent category of wireless networks that operate without fixed infrastructure, offering great flexibility in dynamic environments such as battlefields, remote areas, and smart cities. However, due to their decentralized structure and limited security in routing protocols, MANETs are more vulnerable to attacks compared to traditional wired networks. Specifically, black and gray hole attacks represent a serious threat, as malicious nodes exploit network resources, significantly degrading overall network performance. This thesis presents an efficient attack detection system aimed at enhancing the security of the Ad-hoc On Demand Distance Vector (AODV) routing protocol to detect both black hole and gray hole attacks in MANETs different nodes using deep learning techniques. The research is conducted in two key phases. The first phase involves dataset preparation, where network traffic data is generated through simulations in NS-2 (Network Simulator version 2). These simulations incorporate both normal and malicious behaviors, representing black and gray hole attacks within the context of AODV. After pre-processing and analysis, a refined dataset consisting of 28,691 records with 21 features is extracted from the trace files. In the second phase, the proposed detection system is developed and evaluated. The dataset is divided into training (80%) and testing (20%) subsets. Feature selection is carried out using the Categorical Boosting (CatBoost) importance score, with a threshold applied to select the most relevant features. Additionally, the Synthetic Minority Over sampling Technique (SMOTE) is utilized to ensure balanced data for the model. The system employs advanced deep learning architectures, including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and a hybrid CNN-LSTM model. The experimental results show that the proposed system achieves high detection accuracy across various models: LSTM (98.57%), GRU (98.46%), CNN_LSTM (98.60%), and CNN (98.48%). These results demonstrate hybrid CNN_LSTM model show high accuracy and how well the suggested method works to improve MANET security by precisely identifying black hole and gray hole attacks.</p>", 
  "license": "http://www.opendefinition.org/licenses/cc-by", 
  "creator": [
    {
      "affiliation": "Dilla University", 
      "@type": "Person", 
      "name": "Mesfin  Kiflu"
    }
  ], 
  "headline": "Deep Learning-Based Detection for Black Hole and Gray Hole Attacks in  AODV routing protocol MANETs", 
  "image": "https://zenodo.org/static/img/logos/zenodo-gradient-round.svg", 
  "datePublished": "2025-10-15", 
  "url": "https://nadre.ethernet.edu.et/record/20089", 
  "version": "01", 
  "keywords": [
    "MANET, Deep Learning, XGBoost, CatBoost, Black Hole, Gray Hole."
  ], 
  "@context": "https://schema.org/", 
  "identifier": "https://doi.org/10.20372/nadre:20089", 
  "@id": "https://doi.org/10.20372/nadre:20089", 
  "@type": "ScholarlyArticle", 
  "name": "Deep Learning-Based Detection for Black Hole and Gray Hole Attacks in  AODV routing protocol MANETs"
}