Investigating ChatGPT and cybersecurity: A perspective on topic modeling and sentiment analysis

In early 2023, the Artificial Intelligence (AI) industry experienced a significant advancement with the emergence of OpenAI's ChatGPT, a research product that demonstrated remarkable capabilities and garnered widespread attention. ChatGPT is an advanced chatbot powered by the Generative Pretrained Transformers (GPT) architecture, designed to generate human-like conversations encompassing a wide range of knowledge domains. Many AI researchers are currently engaging with the new technology to understand its functionality and limitations. Various expressions across a range of social media platforms, including Twitter, YouTube, Facebook, and numerous others, are currently under investigation. This research seeks to analyze the opinions of ChatGPT users as it regards cybersecurity. This research is important due to its contribution towards gaining enhanced understanding and devising intricate improvements for the chatbot. The Latent Dirichlet Allocation (LDA) algorithm is utilized to extract relevant topics from the texts. Additionally, to analyze user opinions and decipher the sentiments as either positive, negative, or neutral, we use the Natural language tool kit Valence Aware Dictionary for sEntiment Reasoning (NLTK's VADER) and Robustly Optimized BERT Pretraining Approach (roBERTa) libraries. The data used is obtained from Twitter via the SNScrape library, which aided in the retrieval of over 700,000 tweets via the search terms #chatgptsecurity, #chatgpthackers, #chatgptcybersecurity, and #chatgptcyberthreats. The analysis of the results by the VADER model shows 43.8% positive, 36.3% neutral, and 19.9% negative sentiments. Similarly, the roBERTa model shows 14.1% positive, 53.2% neutral, and 32.7% negative. These results show that there is an ongoing concern about ChatGPT and cybersecurity, especially in malware code generation, hacking, intelligence gathering, and phishing attacks.