A summary of the vision of our most ambitious research-ish project.
Synavate Labs is working on the development of a conceptual Cyber Threat Intelligence Platform known as the “Orion Constellation” However, the repositories appear somewhat confusing due to an excess of creative liberties. An advisor suggested creating a concise summary of the concept and goals. Here’s the TL;DR, with plans to refine the repositories and rename the final product more traditionally if it is not more than a dream.
The Orion Hunter serves as inspiration, not just as a mythological and archetypal figure but also as a stunning constellation in the night sky.
To keep it straightforward, as advised:
Thesis: Adversarial AI is rapidly advancing. Industry literature emphasizes agent swarms trained on hack-the-box and other offensive cyber training courses. These swarms represent sophisticated machine learning systems with reinforcement learning elements, akin to chess and other game-playing agents.
New adversaries necessitate innovative defense strategies.
Orion aims to be an offensive-defensive tool comprising four core components. Note that this remains a research project, and there have been several dead ends and valuable learnings. Much published research on this topic is recent. We’re not cybersecurity experts but seek a team member to assist in system evaluation.
Additionally, we’re developing defensive capabilities to alert the system if it’s under attack—another theoretical challenge even beyond the Orion system itself. One thing for certain is these systems either do or must exist.
Orion AI Threat Intelligence Constellation
Orion Threat Hunters
Cloudflare-based workers (deployed but not yet integrated) monitor CIDR ranges or known threat actors for heuristics indicative of automated system attacks. Understanding these TTPs (Tactics, Techniques, and Procedures) and heuristics is critical. They act as crawlers or scanners and form a network rather than a host-based system. How to dissect the attack vectors and define the “zones of the sky” for monitoring remains unclear.
Orion Agents
A multi-agent network processes data from the hunters, triages the information, and reports it. At this point, humans remain involved, and high-quality signals will be used to train defensive models. Identifying new heuristics for detecting attack vectors is theoretical but promising. Microsoft Research’s Autogen offers purpose-built multi-agent conversational orchestration, with Langchain and others as potential candidates.
Cognitive Synthesis
A synthetic data generator trained on a mix of threats and benign traffic. For different industries, the aim is to develop defensive models capable of withstanding novel attacks.
Orion Network
Orion Network will maintain data encryption in motion and at rest to ensure transaction authenticity. The blockchain system draws inspiration from Dr. Adam Weigold’s research at Cryptic Inc.
Much work remains, and we are exploring less complex projects that can contribute to the system. There are ethical and safety concerns, as these powerful theoretical systems can serve as both weapons and defensive tools.
The underlying concept is to uncover links between seemingly unrelated data—a theory embodied by Synapse, which has multiple use cases.
We’re open to collaboration. Synavate Labs aims to be a micro-research lab evaluating advanced technology for efficacy in contemporary issues. If you’d like to join us on this journey, please reach out.
Synavate Labs
Aylex Riom 