Llama3 with Parameter Efficient Fine-Tuning Approach for 5G Cellular Network Security Text Classification

Third generation partnership project (3GPP) provides unified telecom standards in terms of numerous technical documents related to 5G cellular networks. Understanding these technical documents becomes a time-consuming task. Therefore, researchers use generative artificial intelligence based LLM models to create and understand the 3GPP knowledge base. This helps to improve network securities in terms of threats, incidents, vulnerability, policy compliance, and security intelligence. In this paper, a Llama3 model with parameter efficient fine-tuning (PEFT) approach is presented with the task of 3GPP 5G cellular network security text classification. This approach is compared with multiple transformer models. The 3GPP-based SPEC5G dataset is used for the task of security-related text classification. Our results indicate that the proposed Llama3 with PEFT approach significantly outperforms.