ACTIVE RESEARCH  ·  PHASE 2 SCANNING

AI-Powered
Security Research
Platform

A systematic platform for AI-assisted vulnerability discovery, red-blue adversarial research, and security capability measurement — built on large language models and automated exploit pipelines.

View Findings ↓ GitHub →
scan_phase2.py — GLM-4-flash · Carlini Pipeline A
$ python3 scan_phase2.py linux_bpf
🔍 Target: linux_bpf (325 extracts)
📌 Focus: container escape, eBPF verifier bypass, OOB map access...

[ 1/325] verifier.c_extract_001.c
[ 2/325] verifier.c_extract_002.c 🚨
[ 3/325] syscall.c_extract_001.c
[ 4/325] btf.c_extract_001.c
[ 5/325] core.c_extract_001.c 🚨
...

════════════════════════════════════════════════════
✅ Done: 12 findings, 280 clean, 3 errors
🔴 High-confidence findings (5):
[HIGH] OOB map access @ verifier.c_extract_002.c (conf=85)
[HIGH] register range imprecision @ core.c_extract_001.c (conf=82)
8
Targets Scanned
1,900+
Code Extracts
1
Confirmed CVE Candidate
7
Platform Subsystems
5.3
Highest CVSS Score
// 01 · PLATFORM

Seven-Subsystem Architecture

A full-stack security research platform spanning vulnerability discovery, sandboxed evaluation, supply chain protection, adversarial red-blue engines, and AI capability measurement.

SUBSYSTEM 01

Model Abstraction Layer

Unified SecurityAgent interface decoupled from any single LLM. Supports Claude, GPT, Qwen, DeepSeek, and local models via vLLM/Ollama. Per-experiment token budget and cost tracking.

Claude API OpenAI GLM-4 vLLM Cost Tracking
SUBSYSTEM 02

Vulnerability Discovery Engine

Four scanning pipelines: Carlini-style file scanning (Pipeline A), containerized full-project analysis (B), fuzzer crash triage & exploit development (C), and N-day CVE automation (D).

Pipeline A-D GLM-4-flash Triage Exploit Dev
SUBSYSTEM 03

Sandboxed Evaluation Environment

Three-tier target range: single-container app targets (L1), enterprise multi-segment networks (L2), and OT/ICS critical infrastructure simulation (L3). Full network isolation on all attack containers.

Docker QEMU/KVM Air-gapped L1–L3 Range
SUBSYSTEM 04

Supply Chain Security

Slopsquatting hallucination detection — cross-reference AI-recommended packages against real registries. Behavioral sandbox analysis for newly registered packages. Agent dependency install interception.

Slopsquatting npm / pip Behavior Analysis Install Hook
SUBSYSTEM 05

Agent Behavior Monitor

12-class anomaly detection covering sandbox escape attempts, credential search, deceptive output, evaluator injection, git history tampering, and evaluation-awareness signals. Three-tier alert system.

12 Anomaly Classes Escape Detection 3-Tier Alerts
SUBSYSTEM 06

Red-Blue Adversarial Engine

Autonomous red-team agents (OODA-loop situational awareness, multi-agent squad tactics) vs. blue-team agents (kernel hot-patch via kpatch/eBPF, SDN dynamic network defense, moving target defense). 10 adversarial scenarios.

Red Agent Blue Agent kpatch / eBPF SDN / OpenFlow MTD
SUBSYSTEM 07

Security Capability Tracker

Multi-dimensional benchmark tracking across models: CyberGym score, crash triage accuracy, N-day exploit rate, cost-per-vulnerability. Doubling-time measurement and capability emergence detection.

CyberGym Doubling Time Emergence Detection Model Comparison
View full topology diagram  →

/ end-to-end SVG · 7 subsystems · red-team deep dive · roadmap

Research Findings

Phase 1 scanning covered 8 open-source targets. Below are selected candidate vulnerabilities identified by the GLM-4-flash pipeline and manually triaged. All upstream disclosures follow a 90-day coordinated policy.

CAND-008 · Mosquitto 2.0.21

MQTT 5.0 Property Length Integer Underflow

MEDIUM · 5.3

uint32_t proplen underflow in property__read(): a declared property_length=1 with a 2-byte property causes 0 - 1 = 0xFFFFFFFF, triggering unbounded iteration over all remaining packet bytes — one 48-byte heap allocation per 2 bytes consumed. Exploitable via both CONNECT (pre-auth) and PUBLISH vectors. At 50 concurrent attackers × 5 MB packets: 20× mean / 33× P99 latency increase for legitimate clients.

Disclosed to upstream CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L
CAND-005 · ImageMagick 7.1.2

HeapOverflowSanityCheck Wrong Parameter

HIGH

In TIFF decoder tiling path: HeapOverflowSanityCheck(rows, sizeof(*tile_pixels)) passes rows instead of the full tile size. The check may pass while the subsequent allocation is undersized, leading to an out-of-bounds write during tile pixel aggregation. Triggered by crafted multi-strip TIFF files.

Under analysis ImageMagick · coders/tiff.c
CAND-006/007 · Mosquitto 2.0.21

Persistence File Parsing Underflow

MEDIUM

Two candidate underflow conditions in the Mosquitto persistence file (.db) parsing path, potentially triggering local denial-of-service when the broker restarts and reads a malformed state file. Deeper analysis pending to confirm exploitability and impact scope.

Deep analysis pending Mosquitto · persist/read.c
Phase 2 · Linux Kernel

eBPF / KVM / USB Scan in Progress

CRITICAL · P0

Phase 2 scans targeting Linux 6.12 kernel subsystems: eBPF verifier bounds tracking (325 extracts), KVM x86 instruction emulation (264), and USB HID descriptor parsing (119). High-confidence candidates from the GLM scan are currently undergoing manual triage.

Scanning active kernel/bpf · virt/kvm · drivers/usb
// 03 · METHODOLOGY

Scanning Methodology

Four automated pipelines cover the full spectrum from lightweight per-file scanning to containerized end-to-end exploit development.

A

Carlini File Scanning

Extract 5,000-char segments from C/C++ source files. Feed each segment to an LLM with a security-focused CTF prompt. Confidence threshold at 70%. Rate-limited at 20 s/call to respect API quotas. Results written to JSONL.

B

Containerized Full-Project

Air-gapped Docker container with full source + compiled binary + toolchain. Agent autonomously builds, runs, debugs, and writes findings to /output. Up to 1M token budget per target, 1-hour time limit.

C

Fuzzer Crash Triage

Triage OSS-Fuzz or custom fuzzer output. Agent classifies crashes into 5 tiers (0.0–1.0): from basic crash through controlled crash, info leak, partial control-flow hijack, to full RCE. Develop exploit primitives for tier-4/5 bugs.

D

N-day CVE Automation

Given a list of public CVEs, agent selects exploitable candidates, builds the affected binary version in QEMU, develops an exploit, and reports success rate and cost per successful exploit. Baseline: <$2,000 per Linux kernel exploit.

Phase 2 Target Coverage

Deep scanning of six high-priority attack surfaces representing real-world threat scenarios: container escape, hypervisor escape, USB driver crash, and supply-chain attacks.

# Target Version Threat Scenario Extracts Priority
01 Linux kernel / eBPF 6.12 Container escape via eBPF verifier bypass 325 P0 🔴
02 Linux kernel / KVM 6.12 VM guest-to-host escape via hypercall OOB 264 P0 🔴
03 Linux kernel / USB HID 6.12 Kernel crash on USB device connect (probe RCE) 119 P1 🟠
04 ImageMagick 7.1.2-19 OOB read/write — open file → crash/RCE 514 P1 🟠
05 Poppler 26.04.0 PDF parsing DoS / integer overflow / RCE 359 P2 🟡
06 libxml2 2.13.5 Supply chain: embedded in PHP, Python, Ruby, Node.js ~200 P2 🟡
// OPEN SOURCE

Built in the Open

All scanning infrastructure, triage methodology, and (post-disclosure) vulnerability analysis are published on GitHub under a coordinated disclosure policy. Contributions and collaboration welcome.

View on GitHub → Contact Research Team