Yousa3/Chaitin-Shouyuan-CyberGuard-8B
Chaitin-Shouyuan-CyberGuard-8B by Yousa3 is an 8-billion parameter Qwen3-based large language model fine-tuned for cybersecurity intent moderation. It specializes in classifying user inputs as 'safe' or 'unsafe' based on whether they directly request AI to perform cyberattack operations. With a 32768-token context length, it excels at identifying requests for generating attack payloads, scanning systems, or bypassing security, while distinguishing them from security research or vulnerability remediation discussions.
Loading preview...
Chaitin-Shouyuan-CyberGuard-8B: Cybersecurity Intent Moderation
Chaitin-Shouyuan-CyberGuard-8B is an 8-billion parameter model developed by Yousa3, built upon the Qwen3-8B architecture. Its primary function is to act as a cybersecurity intent moderation assistant, classifying user inputs into safe or unsafe categories. This model is specifically fine-tuned to determine if a user is directly asking the AI to execute cyberattack operations.
Key Capabilities
- Binary Classification: Outputs either
safeorunsafefor user queries. - Cyberattack Detection: Identifies requests for generating attack payloads (e.g., SQL injection, XSS), scanning target systems, bypassing WAFs, writing malware, or structuring penetration tasks.
- Contextual Understanding: Differentiates between malicious intent and legitimate cybersecurity discussions, such as vulnerability repair, code review, or security research, even if sensitive terms are present.
- Built-in System Prompt: Includes a default system prompt for moderation, which can be overridden if needed, guiding the model through a four-step reasoning process (
Intent Analysis,Rule Matching,Evidence Extraction,Final Decision). - Output Format: Provides structured output including a
<think>tag for its reasoning process and a<result>tag for the final classification.
When to Use This Model
This model is ideal for scenarios requiring automated moderation of user-generated content in cybersecurity contexts. It helps in:
- Filtering Malicious Requests: Preventing AI from being prompted to assist in cyberattacks.
- Content Moderation: Ensuring user interactions align with ethical and legal guidelines in security-related platforms.
- Distinguishing Intent: Accurately separating requests for offensive security actions from defensive security practices or academic discussions.
It's important to note that while highly effective, the model serves as an auxiliary intent review tool and should complement, not replace, human review and comprehensive security policies.