AEGIS LLM Security Test Report

6 LLM Comprehensive Defense Evaluation — OWASP Top 10 for LLM Applications

Test Period: 2026-02-19 ~ 2026-02-20 AEGIS v4.0.0 ~ v4.1.0 Scenarios: 25 Attacks + 10 Safe Prompts Models: 6 LLMs

Executive Summary

LLM Self-Defense Rate (Standalone)

LLM + AEGIS Combined Defense Rate

AEGIS Value-Add: Defense Rate Improvement per Model

LLM Security Ranking

Model Comparison

Basic Attacks Defense Rate (15 Scenarios)

Advanced Attacks Defense Rate (10 Scenarios)

Status Distribution per Model (BLOCKED / PARTIAL / VULNERABLE)

Average Response Latency (ms)

False Positive Rate (Safe Prompt Misblock)

Attack Heatmap — Cross-Model View

Detailed Test Results

AEGIS Defense Layer Analysis

Defense Layer Distribution

AEGIS Confidence Distribution

AEGIS Latency (ms) — Per Attack

Safe Prompt Validation (False Positive Check)

10 legitimate prompts tested to verify no false-positive blocking by models or AEGIS.

OWASP Top 10 for LLM — Coverage Map

OWASP Category: Model Defense Coverage