🤖 AI Prompts Cheatsheet
Daniel Osei — AI-Assisted Security Engineer
Why AI Changes the Game for Phishing Detection
The rapid evolution of phishing techniques necessitates a proactive approach to detection. AI tools like ChatGPT and Claude can assist security teams in identifying potential phishing attempts more effectively. By generating scripts, analyzing emails, and automating responses, these AI systems empower analysts to enhance their phishing detection capabilities.
Before You Start: How to Set Context Properly
When engaging AI tools for phishing detection, it’s crucial to set the context clearly. Define your goals: are you analyzing incoming emails, generating phishing simulations, or creating training materials? The clearer your objectives, the better the output from the AI.
Core Prompts Cheatsheet
Weak vs Strong Prompt Examples
Advanced Prompt Techniques
Employing advanced prompt techniques can significantly improve the quality of responses from AI.
- Role Prompting: Assign the AI a specific role. For instance, say, “Act as a cybersecurity expert and draft a phishing training document.” This sets the expectation for output quality and context.
- Chain-of-Thought: Encourage multi-step reasoning by framing your prompts as challenges. Example: “What are the steps needed to investigate a suspected phishing email?” This guides the AI to provide structured responses.
- Few-Shot Examples: Show examples of desired outputs within the prompt. This helps the model understand exactly what you expect.
- Output Formatting: Specify how you want the output presented. For example, “List the characteristics of phishing emails in bullet points.” This adds clarity and aids in further usage.
Claude vs ChatGPT: Which Works Better For This
While both Claude and ChatGPT have unique strengths, Claude may offer better contextual understanding due to its structured training. However, ChatGPT tends to produce more human-like phrasing, which can be beneficial for drafting content. Experiment with both to identify which aligns best with your needs!
Tips for Getting Consistent Results
To ensure reliable outputs, consider the following:
- Context Setting: Always preface your prompts with additional context to guide the AI effectively.
- Specificity: Avoid vague instructions. The more specific and tailored your prompt, the more relevant the responses will be.
- Iterative Refinement: Don’t hesitate to refine your prompts based on previous outputs. This can lead to increasingly better results.
Quick Reference: All Prompts in One Place
Here’s a quick reference of all the prompts discussed:
- Analyze the following email for phishing indicators: [PASTE EMAIL TEXT]
- What are the main characteristics of this phishing email? [PASTE EMAIL TEXT]
- Generate a phishing simulation email aimed at corporate employees.
- Provide a list of best practices for detecting phishing emails.
- Help me create a phishing response playbook.
- What are some recent phishing trends to be aware of for 2023?
- Draft a phishing awareness announcement.
- List tools and software for detecting phishing attempts.
- Summarize key features for an anti-phishing tool.
- What response strategies should be in place following a successful phishing attack?