Social Engineering Attacks Using Technical Job Interviews: Real-life Case Analysis and AI-assisted Mitigation Proposals — (Supplementary Material)

This supplementary material provides the complete set of prompts and responses used to evaluate ten large language models (LLMs) in detecting security risks within an obfuscated JavaScript snippet, according to a published study (10.3390/info17010098). The code snippet was analyzed through three prompt types: an operative prompt for pre‑execution safety checks, a simple prompt for basic risk assessment, and a technical prompt for detailed inspection of obfuscation and remote code execution. Responses were obtained through official web interfaces and third‑party services, including ChatGPT, Grok, Claude, Lumo, Perplexity and HuggingChat, using secure HTTPS connections and WebSocket streaming to replicate real‑world interaction conditions. For each model, the full outputs are included to ensure transparency and reproducibility of the experiment.