Scholarlink

<p>The integration of artificial intelligence (AI) and machine learning (ML) within the Open Radio Access Network (O-RAN) architecture introduces enhanced automation and anomaly detection capabilities. However, this openness also increases exposure to security threats, particularly backdoor attacks, which compromise AI-driven xApps and threaten network integrity. In this work, we propose HybxLSTM, a hybrid convolutional Extended Long Short-Term Memory (xLSTM)-XGBoost model designed to enhance anomaly detection and mitigate backdoor attacks in O-RAN environments. Our model leverages Extended Long Short-Term Memory (xLSTM)’s ability to capture long-term dependencies in network traffic, convolutional layers for efficient feature extraction, and XGBoost for robust classification and interpretability. The proposed solution ensures high detection accuracy while maintaining computational efficiency for real-time applications within the Near-Real-Time RAN Intelligent Controller (Near-RT RIC). Extensive evaluations on the TON IoT dataset demonstrate the effectiveness of HybxLSTM, achieving near-perfect accuracy, precision, and recall while maintaining a low false positive rate and attack success rate. Our approach outperforms traditional LSTM and BiLSTM models, addressing key security challenges in AI-powered O-RAN systems. This work underscores the need for adaptive, scalable anomaly detection mechanisms to secure AI-driven telecommunications infrastructure against emerging threats.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybxlstm: a hybrid xLSTM-XGBoost model for securing O-RAN against backdoor attacks

  • 期刊论文

  • 发表日期：

    2025年9月26日

查看全文
  • Rouaa Naim,
  • Hams Gelban,
  • Ahmed Badawy

摘要

The integration of artificial intelligence (AI) and machine learning (ML) within the Open Radio Access Network (O-RAN) architecture introduces enhanced automation and anomaly detection capabilities. However, this openness also increases exposure to security threats, particularly backdoor attacks, which compromise AI-driven xApps and threaten network integrity. In this work, we propose HybxLSTM, a hybrid convolutional Extended Long Short-Term Memory (xLSTM)-XGBoost model designed to enhance anomaly detection and mitigate backdoor attacks in O-RAN environments. Our model leverages Extended Long Short-Term Memory (xLSTM)’s ability to capture long-term dependencies in network traffic, convolutional layers for efficient feature extraction, and XGBoost for robust classification and interpretability. The proposed solution ensures high detection accuracy while maintaining computational efficiency for real-time applications within the Near-Real-Time RAN Intelligent Controller (Near-RT RIC). Extensive evaluations on the TON IoT dataset demonstrate the effectiveness of HybxLSTM, achieving near-perfect accuracy, precision, and recall while maintaining a low false positive rate and attack success rate. Our approach outperforms traditional LSTM and BiLSTM models, addressing key security challenges in AI-powered O-RAN systems. This work underscores the need for adaptive, scalable anomaly detection mechanisms to secure AI-driven telecommunications infrastructure against emerging threats.