Cybersecurity R&D

Pioneering the next generation of defensive and offensive cyber capabilities under Project Anti-Matter.

Singularity Shield: Autonomous XDR

Static rulebooks and basic heuristics are failing. The future of security operations isn't just about detecting breaches; it's about real-time, zero-human-latency mitigation.

The Objective

To build an agentic, behavioral threat detection engine powered by local, fine-tuned LLMs. This tool will continually monitor system logs, active API endpoints, and raw network telemetry to crush unfolding attacks (like multi-stage ransomware) within milliseconds.

The Tech Focus

  • Behavioral Agentic Detection: Moving away from traditional IOC signatures to true anomaly-based reasoning.
  • Self-Healing Orchestration: Teaching autonomous AI agents to safely isolate compromised server containers, revoke tokens, and rewrite firewall rules mid-attack.
  • Low-Latency Pipelines: Keeping inference speeds fast enough to intercept automated scripts.

How to Get Involved

We are looking for Systems Engineers, Linux Kernel Specialists, and AI Engineers who want to experiment with agentic action-spaces.

Void Unpacker: AMAP (AI Malware Analysis Platform)

Attackers are already leveraging AI to build polymorphic malware that mutates code mid-flight. Defensive tools have to evolve.

The Objective

We are building a highly specialized deep-analysis sandbox that uses an LLM trained explicitly on assembly code, binary syntax, and reverse-engineering logic. The tool acts as an autonomous decompiler that can safely execute, force-unpack, and reverse-engineer AI-obfuscated malware.

The Tech Focus

  • Automated Decompilation Assistant: AI that automatically reads raw assembly, maps out obfuscated execution paths, and documents structural blueprints.
  • Safe Behavioral Micro-Sandboxing: Orchestrating dynamic, isolated sandbox environments to force malware to drop its payloads.
  • Predictive Execution Modeling: Analyzing encrypted payloads before they run to guess their command-and-control (C2) blueprints.

How to Get Involved

If you live in Ghidra, IDA Pro, x64dbg, or love Reverse Engineering and Assembly, this project is for you. We need developers who know how to safely dissect binaries and want to teach AI to do it at scale.

Exosphere Mirage: HORNET (OSINT & Active Deception)

Defending a network perimeter passively isn't enough. This project flips the script, combining OSINT profiling with hyper-realistic active deception to turn the tables on threat actors.

The Objective

To create a two-part defensive suite. Part one maps your network as an attacker sees it. Part two uses that data to instantly generate a custom, LLM-driven honeypot that perfectly mirrors those vulnerabilities—trapping threat actors to extract their TTPs.

The Tech Focus

  • Autonomous Attack Surface Mapping: Scraping public data to build a dynamic OSINT profile of your network.
  • Generative Honeypots: AI that constructs realistic terminal environments and fake databases on the fly to match what a hacker is searching for.
  • Attacker Profiling: Using an active LLM agent to string along a hacker inside the honey-network to capture their digital signatures and toolkits.

How to Get Involved

Calling all Network Engineers, OSINT Specialists, Full-Stack Devs (Next.js/Tailwind), and Deception Researchers. We need people who know how network protocols tick and understand how to construct convincing traps.

Join the Research Lab

These projects are ambitious, and we're building a community of passionate engineers to bring them to life. If you want to contribute, learn, and build the future of cybersecurity, our lab is the place to be.

Join our Discord