Rat5ak/CORUNA_TECHNICAL_ANALYSIS
Security & PrivacyA static reverse-engineering teardown of the Coruna iOS/macOS Safari watering-hole exploit chain, written for threat analysts and iOS security researchers.
What this is
This repository is a published threat-intelligence paper, not a tool. It documents a full static reverse engineering of Coruna, a multi-stage exploit chain attributed to UNC6691 that targets Safari/WebKit and the XNU kernel on arm64e iOS and macOS devices readme_abstract. The chain runs from browser exploitation, through Mach-O parsing and PAC bypass, to kernel compromise via IOSurfaceRoot (CVE-2023-41974) and a read-write remount of the root filesystem.
The analysis covers roughly 700 KB of obfuscated JavaScript across 28 files and 15 unique modules, plus a 2 MB ARM64 kernel exploit DYLIB (dump.bin, 649 functions). It includes XOR string decryption (191 unique strings across 1,250 instances), embedded WebAssembly module extraction, ARM64 instruction-mask mapping, kernel import categorization (265 imports), and kill-chain integration readme_abstract.
Who it is for
Malware analysts, detection engineers, and iOS/macOS security researchers who want to understand how a state-grade WebKit exploit is obfuscated, how it locates symbols in the dyld shared cache from JavaScript, and how it defeats PAC without forging a signature. Readers should be comfortable with JavaScript, Mach-O structure, ARM64 assembly, and exploit terminology; the paper assumes that background rather than teaching it.
How to use it
There is no build, install, or runnable component. Read the paper in order — it follows the actual attack chain:
- Infrastructure and C2: the single hardcoded delivery domain, payload inventory, XHR-based beacon protocol, and DOM-injection stealth readme_infra.
- Obfuscation layer: the
globalThis.vKTo89module namespace,OLdwIx/tI4mjAimport and registration, XOR string and numeric-constant encoding, and anti-analysis patterns such as empty error messages readme_obfuscation. - Mach-O parser and dyld cache walker, then WebKit exploit primitives, trigger mechanisms, PAC bypass, JIT cage escape, and final payload assembly.
- Appendices: decoded string inventory, file hashes, YARA and Suricata rules, and behavioral indicators for defenders readme_iocs.
Detection teams can use the indicators-of-compromise section directly; the network detection rules reference the observed delivery domain readme_domain.
Constraints and caveats
- The document is a point-in-time analysis dated March 2026; the author acknowledges that all 13 payloads remained live and byte-identical at publication, and that the infrastructure later moved behind CloudFront readme_opsec.
- Findings are static analysis. Runtime behavior is inferred from decoded code rather than demonstrated, and the kernel-stage conclusions depend on one recovered
dump.bin. - Attribution to UNC6691 and the additional delivery domains are drawn from Google Threat Intelligence Group reporting, not independently verified in the repository readme_gtig.
- No license is declared, so reuse of the text or the YARA/Suricata rules has no explicit grant.
- The paper contains descriptions of exploit internals. It is reference material for defenders; it does not include a runnable exploit, and none should be reconstructed from it.
HotRepo's take
The repository has no code language and no HotRepo ranking or star-curve history, which fits a static paper rather than a software project. Its value lies in being one of the few public English-language deep dives that decodes an entire obfuscated exploit framework end to end — from the vKTo89 module namespace to a PAC bypass built on GOT swaps — instead of stopping at IOCs. The most reusable parts for practitioners are the decoded string inventory and the detection rules in Appendix B, while the attribution and infrastructure timeline should be read as GTIG-derived context rather than independently confirmed facts.
Written by AI from the project's public information and HotRepo's records, updated . The project's own docs are the source of truth.
You might also like
- khanhduytran0/corunaDeobfuscated iOS exploit chain toolkit for security researchers studying captured payloads.
- AloneMonkey/frida-ios-dumpExtract a decrypted IPA from a jailbroken iOS device over Frida for reverse-engineering and analysis.
- cyberkaida/reverse-engineering-assistantGhidra extension exposing reverse-engineering tools to LLM clients through an MCP server, for interactive or headless binary analysis.
- Minglink/dsh-infinite-gen-5A reverse-engineering workbench inside DeepSeek Harness for authorized malware and protocol analysis, bundling static, dynamic and traffic evidence in one workflow.
- skylot/jadxTurns Android APKs and Dalvik bytecode back into readable Java for reverse engineers and security analysts.