ArgusAI Security Detection Benchmark
ArgusAI maintains an automated security test corpus with mutation fuzzing across eight obfuscation techniques. Latest run: 7,480 tests at 100% detection rate with zero false positives.
ArgusAI maintains an automated security test corpus with mutation fuzzing across eight obfuscation techniques. Latest run: 7,480 tests at 100% detection rate with zero false positives.
- Baseline tests
- 920
- With mutations
- 7,480
- Detection rate
- 100%
- False negatives
- 0
- False positives
- 0
Methodology
Tests cover prompt injection, jailbreak, SQLi, XSS, resource exhaustion, secrets extraction, MCP abuse, and safe prompts. The mutation engine applies Base64, Hex, Unicode zero-width, HTML entities, markdown, ROT13, random spacing, and homoglyphs.