Validating AI Findings

We seek your expertise to validate findings from our AI model, as your contributions are invaluable in improving their accuracy and reliability. By participating, you can help refine insights, identify potential errors, and enhance the credibility of our AI system. In turn, this will enable you to discover issues more efficiently and address specific concerns more quickly. For successful validations auditors receive credits that can be converted to $AUDIT tokens later on.

What have we done?

The development is focused on the conversion of code snippets into abstract syntax trees (ASTs) and analyzing them with Graph Neural Networks (GNN). In addition, ensemble methods are used to improve recognition accuracy and eliminate errors, in particular false positives, which often occur in large ASTs. A LLM-based examination assistant uses Retrieval Augmented Generation (RAG) techniques, code visualizations with Slither Printer Graph and fine-tuning techniques which prepares the ASTs in a comprehensible presentation to enable auditors to specifically check and validate vulnerabilities.

How to Validate AI Findings?

Step 1: Login into the AuditOne app (https://app.auditone.io/).

Step 2: Go to the tools section.

Step 3: Click on “Validate findings” in the audit assistant section.

Step 4: Start by Selecting an Issue.

Step 5: Review the issue details carefully to understand what is at stake.

Step 6: Examine the code pertaining to the issue. If you find something, document it carefully.

Step 7: Verify the Issue. Ensure the issue is valid, effectively approving it or not.

How Is It Judged?

• Triagers evaluate every reported issue, whether by one person or many, and assign - 10 credits for valid findings.

• If 4 out of 5 people confirm an issue is real, those 4 earn 10 credits each.

• You can retain your credits to continue using our tools for free once we transition to a monetized model, or you can exchange them for $AUDIT tokens at a conversion rate of 1 Credit = 5 $AUDIT tokens.