exploits - Security research and exploit development: vulnera

项目说明

Security Research & Exploit De

A general-purpose security research repository spanning browser exploit chains, post-exploitation tradecraft, identity and cloud attacks, and the assessment work that surrounds them. The contents are organized into three broad bodies of work:

C — patch analysis, triggers, and full chains for Chrome and Firefox bugs from 2024 through 2026.

Tooling — a catalog of offensi (C2, lateral mo, AD CS, cloud identity, e, LLM attacks, kernel LPE, and more), each paired with detection guidance.

Assessment deli — written reports for stakeholders, including a Databricks Apps platform review aimed at security leadership.

Every offensi ships with a defender-side counterpart. The repository is meant to feed scheduled enterprise log-scan agents and red-team tabletop work, so breadth is intentional.

---

Purpose & Disclaimer

This repository exists to support internal security education and risk awareness. The exploit reproductions, attack chain demonstrations, and assessment tooling are built as red-team exercises that show stakeholders — including non-technical business users — what can go wrong when production systems lack adequate human o and go controls.

All materials are for educational purposes and authorized security testing only. Research was conducted under explicit written authorization. Exploit code is provided to support defensi, reproduction, and security assessment. Do not use any code or technique against systems you do not own or ha written authorization to test. See SECURITY.md for the full responsible-disclosure policy and containment architecture.

Safeguards in place:

Every CVE here targets an already-patched browser with a publicly a PoC. No zero-days are published. The contribution is in the chaining, primitives, and delivery infrastructure built around them.

All tool networking is loopback-only (127.0.0.0/8 or the Docker bridge). This is enforced in code by ContainmentGuard — not by con.

The Docker Compose lab runs on an internal: true network with no internet gateway.

The beacon refuses to run its exec command outside of a Docker container, and every other beacon command is on a hardcoded allowlist.

The IDOL persistence demo (--plant) only installs a harmless cron heartbeat that --cleanup removes.

The GitHub Pages dashboard runs entirely on simulated data — there is no live C2 connectivity in the public site.

---

Contained Lab En

A Docker Compose lab provides a fully isolated en for running tools end-to-end. All services run on an internal Docker network with no internet access, so tools that "phone home" simply talk to mock services running alongside them.

The lab can be brought up incrementally — the base make lab-up is enough for most C2 and exploit-server work, and additional mock IdPs and OAuth issuers come up targets so you only run what your scenario needs.

| Service | Port | Description | How to start |
|---------|------|-------------|--------------|
| C2 server | 127.0.0.1:8443 | Operator API + beacon protocol | make lab-up |
| Exploit server | 127.0.0.1:9090 | Serves C, recei | make lab-up |
| Target app 1 | 127.0.0.1:8501 | Simulated Databricks Streamlit app | make lab-up |
| Target app 2 | 127.0.0.1:8502 | Second target for lateral mo | make lab-up |
| Mock Entra IdP | 127.0.0.1:9100 | Device code, token, PRT SSO endpoints | make lab-up |
| Mock IMDS | 127.0.0.1:9200 | AWS / GCP / Azure metadata service mock | make lab-up |
| LLM copilot app | 127.0.0.1:8080 | Ollama-backed enterprise copilot (injection target) | make lab-llm-up |
| Mock OIDC issuer | 127.0.0.1:9300 | GitHub Actions OIDC simulation (WIF abuse) | make lab-oidc-up |
| Mock SAML SP/IdP | 127.0.0.1:9400 | SAML assertion target (Golden SAML demo) | make lab-saml-up |
| Mock Databricks | 127.0.0.1:9500 | Databricks Apps OAuth/OBO mock | make lab-databricks-up |
| AD CS lab | 192.168.56.10 | Windows DC + Enterprise CA (, host-only) | make lab-adcs-up |

Containment. tools/lib/containment.py — and its Rust counterpart in tools/rust/containment/ — is the trust boundary. It enforces loopback-only networking, non-root execution, tmpdir isolation, and Docker en detection across every tool. If a tool tries to step outside the sandbox (talk to a real tenant, reach a real metadata service, write outside the fixture root), the guard refuses before any side effect happens.

---

Tools

Each tool below ships under tools/ and has a sibling detection/ directory containing Sigma rules, KQL queries, or other defender-side guidance for the same technique. The brief descriptions are intentionally compact — each linked README expands on usage, lab requirements, and detection.

C2 & Infrastructure

C2 Server & Beacon — tools/c2/. Modular C2 with pluggable transports (HTTP polling, WebSocket, gRPC, SMB / Unix pipe, DNS-o, HTTP/3, DoQ, ology. Flask server with X25519 + ChaCha20-Poly1305 session crypto, hardcoded command allowlist, and loopback-only enforcement.

C2 Transports — tools/c2/transports/. One subdirectory per transport (http_polling/, websocket/, grpc/, passi/, dns_o/, http3/ for QUIC, doq/ for DNS-o, and teams-graph/ mocks). Each ships with Sigma / KQL detection alongside the offensive code.

C2 Framework Profiles — tools/c2/profiles/framework-mimics/. Transport-signature profiles that mimic AdaptixC2, emp3r0r, APEX, Mythic, Havoc, Brute Ratel C4, and NightHawk. Use them with the beacon to detection rules against specific framework footprints.

C2 Relay — tools/c2/relay/. P2P relay node that supports beacon chains of depth ≥2, for testing detections that key off intermediate hops rather than direct callbacks.

Dashboard — tools/dashboard/. Session management console with multi-transport session view, profile editor, and a relay topology graph.

Acti & Kerberos

AD CS Abuse — tools/ad-cs/. The full ESC1 through ESC16 toolkit: an LDAP-based template enumerator following Certipy patterns, per-ESC exploit modules, a chain orchestrator, and the Shadow Credentials 2026 update co January 2026 NGC key-write hardening. Every module is gated to the corp.lab.local lab domain.

Kerberos Lateral Mo — tools/kerberos/. S4U2self / S4U2proxy abuse, RBCD chain, NTLM relay with EPA recon (RelayInformer-class), the C NTLM reflection LPE, AES-only Kerberoasting (post C RC4 enforcement), and ser triage tooling.

Lateral Mo

Lateral Mo — tools/lateral-mo/. RPC-based DCOM / TSCH / SCMR / WMI execution; SCCM ELEVATE1/2 plus the TAKEOVER-5 Entra-integration chain (SpecterOps, No 2025); Azure Arc MSI pivot with C (himds pipe DACL); Exchange hybrid evoSTS token forge.

Snowflake Pivot — tools/lateral-mo/snowflake-pivot/. Snowflake-specific lateral primitives: Storage Integration enumeration (Chain E), Direct Share + replication-group exfil (Chain G, audit-bypass data motion), SPCS egress matrix (Chain H), Polaris / Iceberg catalog pivot (Chain K), UDF EAI breakout (Chain M), SPCS base-image posture probe (Chain H supply-chain extension), bind-parameter e against QUERY_HISTORY. Talks to mock-snowflake on 9600.

Cloud Identity

Cloud Identity Attacks — tools/cloud-identity/. The modern cloud-identity surface: Workload Identity Federation wildcard sub abuse, Golden SAML, Silver SAML (secondary cert), SyncJacking via ImmutableId takeover, E de 2026 PhaaS (Broker client ID FOCI path), FOCI Conditional Access bypass, PRT extraction via de, and a CloudTrail-blinding catalog. Talks to the lab mocks: mock-oidc (9300), mock-saml (9400), mock-entra (9100/9102).

Snowflake Cloud Identity — tools/cloud-identity/snowflake/. Snowflake-specific identity abuse for the post-UNC5537 / post-MFA control surface: JWT key-pair signer (Chain F — ser key theft from CI / orchestration hosts), PAT scope walk, SCIM token har with a role-race primiti audit (Chain J), external OAuth scope-drift audit (Chain L). Talks to mock-snowflake on 9600.

Entra ID Abuse (legacy) — tools/entra-abuse/. Earlier de phishing, PRT simulation, and token-replay work. Kept for historical reference; current Entra work li cloud-identity/.

Kernel LPE (Windows)

Kernel LPE — tools/kernel-lpe/. Research notes and harnesses for the AFD.sys pool-grooming family (C / -20831 / -20860 / -21241), the CLFS buffer o C, and the I/O Ring arbitrary-R/W primiti. Includes pool feng-shui and NT-structure reference docs. All of this is gated behind EXPLOIT_LAB_KERNEL=1 and a kern-lab-* hostname, by safety_harness.py before anything dangerous runs.

Phishing & Initial Access

AiTM Kits — tools/phishing/aitm-kits/. Simulators for the major ad phishing kits — Tycoon2FA (against mock-aitm-proxy 8100 / mock-m365-login 8101), Sneaky2FA's browser-in-browser , and Rockstar2FA — plus reference docs for E. Importantly, this section also documents the defender controls that actually break these kits: passkey mitigation and Conditional Access token protection.

Social Engineering — tools/phishing/social-engineering/. The "Fix" delivery family: ClickFix (clipboard → Run dialog), FileFix (Explorer address-bar primitive), and ConsentFix (OAuth-consent bypass simulated against mock-entra). Tracks how these chains e from 2023 through 2026.

Passkey Phish-Resistance — tools/phishing/passkey-resistance/. WebAuthn origin-binding fundamentals, a phish-resistance matrix that scores common attack types against MFA factors, the downgrade attack surface, and the cross-device QR-relay risk that has emerged with passkey rollouts.

Tabletop — tools/phishing//. Tabletop scripts for CFO-impersonation and Scattered-Spider helpdesk scenarios. Discussion-only — no voice synthesis is included.

Supply Chain

GitHub Actions — tools/supply-chain/github-actions/. pull_request_target ("pwn_request") simulation against a mock on 8120, the UNC6426 OIDC chain (mock-oidc 9300 → mock-imds 9200), self-hosted-runner persistence, and the tj-actions-class supply-chain compromise pattern.

npm Worm — tools/supply-chain/shai-hulud-class/. Shai-Hulud-class postinstall har worm simulator running against a mock npm registry on 8110.

PyPI .pth Stealer — tools/supply-chain/pypi-pth-class/. LiteLLM-style .pth persistence simulator with a runtime-detector defender tool, exercised against mock-pypi on 8111.

E (Rust)

HW-BP Syscalls — tools/rust/syscalls-hwbp/. Hardware-breakpoint syscall dispatch that bypasses userland EDR hooks. Windows-specific.

Modern Sleep Masks — tools/rust/sleep-mask-modern/. Cronos (fiber + RC4), RustyCronos, and HWBP-dri .

Threadless Injection — tools/rust/threadless-inject/. Module stomping, Transactional NTFS, and DLL-notification-callback hijack approaches.

ETW-TI Awareness — tools/rust/etw-ti-aware/. ETW provider enumeration keyed on EDR GUIDs — useful for understanding what telemetry the host is feeding into ETW-TI.

Call Stack Spoofing — tools/rust/callstack-spoof/. SilentMoonwalk-pattern gadget finder plus a with_spoofed_stack() RAII wrapper for hiding suspicious frames.

Call Stack Detection — tools/rust/callstack-detect/. The defender-side pair of the spoof crate: a kernel-ETW call-stack collector and unwind , with a spoof- matrix showing which spoof patterns survi detections.

Patchless AMSI / ETW Bypass — tools/rust/amsi-patchless/. HWBP (DR0 / DR1) arm and disarm primitives. Both the SetThreadContext and NtContinue are documented.

BOF / COFF Loader — tools/rust/bof-loader/. Goblin-based COFF parser, symbol allowlist, and OutputSandbox capture. See also tools/bofs/ for safe BOF implementations that exercise the loader.

BYO — tools/byovd/. Hash-only manifest, H checker, LOLDrivers sync (hash-only — no dri are committed), H enumeration, and the EDR-killer class targeting the broader 2026 BYO.

LLM & Agent Attacks

M365 Copilot — tools/llm-attacks/m365-copilot/. EchoLeak (C, zero-click email → Copilot → exfil) and ShareLeak (C, Copilot Studio form-field injection). Both run against mock-copilot on 8090.

Agentforce — tools/llm-attacks/agentforce/. PipeLeak public lead-form hijack simulation against mock-agentforce on 8091.

MCP Abuse — tools/llm-attacks/mcp-abuse/. Tool poisoning and capability-confusion patterns, plus the Git MCP server C (C / -68144 / -68145 — commit-message injection, diff injection, path tra).

Cortex — tools/llm-attacks/cortex/. Cortex Search index-poisoning bench (rank hijack, payload injection, semantic-model confusion); Cortex Agent + MCP poisoning bench (Chain I — second-order tool calls and agent-executed SQL from poisoned tool output); Cortex Guardrails FP/FN test harness with a structurally-deri payload corpus. Talks to mock-snowflake on 9600 and mock-snowflake-mcp on 9620.

E — tools/llm-attacks/eval/. AgentDojo-format harness paired with a PromptArmor adapter (instruction hierarchy + cosine similarity), an MCPSec adapter (schema + capability enforcement), and a defense-benchmark runner.

Indirect Injection Corpus — tools/llm-attacks/indirect-injection/. Prompt-injection payloads organized by delivery channel — PDF, DOCX, HTML, email (plaintext and HTML), calendar invites, and image alt text — including a m365_copilot channel for the Copilot-specific surface.

Browser

Browser Extension Supply-Chain — tools/browser-ext-attacks/. MV3 lab catalog: cookie theft, session hijack, form-grab, DNR redirect abuse, and the Cyberha update-hijack simulation.

Browser-Nati — tools/browser-nati/. WASM post-exploitation payload, with delivery via MV3 extension, service worker, or XSS.

Exploit Framework — tools/framework/. Equation Group–inspired exploit orchestration glue.

Fuzzing — tools/fuzzing/. JIT (GVN, LICM, Range Analysis), IPC, and V8 Turbofan fuzzers.

Kubernetes & Cloud-Native

Kubernetes Post-Exploitation — tools/kubernetes-postex/. Runc container breakout (C / -52565 / -52881) chained to kubelet, the NodeRestriction admission bypass (C / -5187), and the AKS C work (CVSS 10, Azure CNI IPAM socket). Gated behind EXPLOIT_LAB_K8S=1.

Linux Post-Exploitation

eBPF Rootkit — tools/linux-postex/ebpf-rootkit/. LinkPro-class XDP / TC + magic-packet C2 analysis, a Boopkit / TripleCross / ebpfkit corpus comparison, and a Rust simulation stub. Gated behind EXPLOIT_LAB_EBPF=1.

macOS / Mobile / Firmware (Research)

macOS — tools/macos-postex/. TCC bypass research: C and an entitlement-o catalog cataloging real applications that ship with TCC-relevant permissions; SLAP / FLOP side-channel analysis; SysBumps KASLR break. Documents only — no PoC.

Mobile Landscape — docs/analysis/mobile-landscape-2026/. iOS DarkSword chain, Coruna kit, Pixel 9 zero-click, Pwn2Own Ireland 2025, and Lockdown Mode effecti. Research-only.

Firmware — docs/analysis/firmware-landscape-2026/. Hydroph0bia (Insyde H2O SMM), LogoFAIL successors, the June 2026 UEFI cert expiry, and a defender-side in.

Persistence

RMM Abuse — tools/persistence/rmm-abuse/. ScreenConnect / AnyDesk / QuickAssist attacker patterns with persistence mechanics and detection. Cross-tool kill-chain correlation makes the case that an RMM install combined with BYOVD is a strong pre-ransomware indicator.

ESXi Post-Exploitation — tools/persistence/esxi-postex/. VM kill chain, ESXi lifecycle post-ex, and Nutanix AH.

Legacy & Support

IDOL — tools/idol/. Lateral-mo PoC co credential harvest, persistence, and C2 beaconing.

Rust Target Tools — tools/rust/. The Rust workspace. Build with cd tools/rust && cargo build --release.

Post-Exploit Staging — tools/post-exploit-staging/. Three-tier staging architecture.

Forensic Analysis — tools/forensic-analysis/. Artifact detection and audit-gap analysis.

— tools//. Pre-exploitation browser fingerprinting.

---

C

The repository tracks Chrome and Firefox CVEs from 2024 through 2026 — patch analysis, triggers, and complete chains. The full table is collapsed below; for the canonical index with implementation status, see c.md.

<details>
<summary>C across Chrome and Firefox (2024–2026) — click to expand</summary>

| CVE | Target | Year | Technique | Level | Path |
|-----|--------|------|-----------|-------|------|
| C | Chrome V8 Maglev | 2024 | OOB Write | ACE | c/2024/C/ |
| C | Chrome V8 Wasm S128 | 2024 | Type Confusion | ACE | c/2024/C/ |
| C | Chrome V8 Object Transitions | 2024 | Type Confusion | ACE | c/2024/C/ |
| C | Chrome Wasm JSPI | 2025 | Sandbox Escape | ACE | c/2025/C/ |
| C | Chrome ANGLE WebGL2 | 2025 | UAF | UAF | c/2025/C/ |
| C | Chrome V8 Property Array | 2025 | Type Confusion | ARW | c/2025/C/ |
| C | Chrome CSS FontFeature | 2026 | UAF | UAF | c/2026/C/ |
| C | Chrome Skia Glyph Atlas | 2026 | OOB Write | OOB | c/2026/C/ |
| C | Firefox SpiderMonkey | 2024 | Type Confusion | Trigger | c/2024/C/ |
| C | Firefox AnimationTimeline | 2024 | UAF | UAF | c/2024/C/ |
| C | Firefox JIT Range Analysis | 2024 | BCE | ARW | c/2024/C/ |
| C | Firefox Pri JS | 2024 | Sandbox Escape | ACE | c/2024/C/ |
| C | Firefox IPC | 2025 | Sandbox Escape | Trigger | c/2025/C/ |
| C | Firefox Promise | 2025 | OOB | OOB | c/2025/C/ |
| C | Firefox IonMonkey BCE | 2025 | OOB | ARW | c/2025/C/ |
| C | Firefox Wasm GC | 2026 | UAF | ACE | c/2026/C/ |
| C | Firefox Wasm JIT | 2026 | Type Confusion | Trigger | c/2026/C/ |

Level key: ACE = arbitrary code execution, ARW = arbitrary read/write, UAF = use-after-free demonstrated, OOB = out-of-bounds access, Trigger = bug trigger only.

</details>

---

Directory Structure

For more granular indexes, see tools/, docs/analysis/, and docs/methodology/ — each has its own README.

---

Getting Started

1. Clone the repository and install lab dependencies: pip install -r requirements-lab.txt
2. Bring up the contained lab (requires Docker): make lab-up. Use make lab-status to confirm e is healthy and make lab-down when you are finished.
3. To run tools locally without the full lab, start the C2 server in one terminal (python3 tools/c2/server.py) and a beacon in another (python3 tools/c2/beacon/beacon_client.py).
4. Browse cves/ for specific C. Each C contains its own README with setup instructions.
5. For the Databricks assessment dashboard, see the section below.

---

Databricks Apps Assessment

A secondary deli in this repository is an interacti assessment of the Databricks Apps platform, prepared for security leadership. It lives at reports/databricks-apps-assessment/ and is rendered as a Streamlit dashboard.

Run it. cd reports/databricks-apps-assessment && pip install -r requirements.txt && streamlit run app.py

Build flow. Source files li src/ (_00_header.py … _99_dispatch.py); python build.py concatenates them into the deployable app.py. Edit the src/ files, never app.py directly. python build.py --check the build is up to date.

Findings the report covers include OAuth On-Behalf-Of (OBO) token-abuse paths in Databricks Apps, the Databricks app identity chain and ser exposure, detection gaps in the Databricks audit logs, and concrete recommendations for monitoring, go, and access-control impro.

---

Links

Repository: github.com/AndrewAltimit/exploits

GitHub Pages: andrewaltimit.github.io/exploits

Li: andrewaltimit.github.io/exploits/dashboard — runs in-browser

License

Dual-licensed under Unlicense and MIT.

项目信息

开发语言：Python

Star 数：13

技术要点

该项目属于防护思路方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕Security research and exploit development: vulnerability analysis, exploit chain implementation, post-exploitation tradecraft, and defensive assessment tooling. Covers browser engines, persistence mechanisms, credential harvesting, C2 patterns, and AI-accelerated attack automation.展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。

登录后才能参与讨论

下载链接

会员才能下载

登录后开通会员

同分类推荐

暂无同类资源

资源标题	exploits - Security research and exploit development: vulnera
所属分类	防护思路
发布作者	admin
发布时间	2026-10-06 00:11:29
下载权限	免费下载
附件数量	2 个下载链接
浏览次数	155
讨论数量	0
主题状态	正常讨论中