We Build Tools That
Fight Back
Oubliette Security is a veteran-owned cybersecurity company specializing in AI security, cyber deception, and threat intelligence. Our roots are in purple teaming — bridging offense and defense to build security that gets smarter with every attack.
Our Mission
The AI security market is where web application security was in 2005. Everyone knows there's a problem, but most solutions offer only passive detection — block the attack, log the event, move on.
That approach gives attackers exactly what they need: instant feedback to iterate. Coming from a purple team background, we believe in a fundamentally different approach: make attackers think they succeeded. Feed them fake data. Waste their time. Generate intelligence from their techniques.
Every AI system deserves a firewall that fights back. That's why our core library is open source under Apache 2.0 — because small teams and solo developers deserve the same protection as enterprises.
Our Journey
Problem identified
After years of watching AI systems fall to prompt injection in real-world deployments, we knew the "block and notify" approach was fundamentally broken. Attackers iterate too fast.
Core detection built
5-stage tiered pipeline achieving 85-90% detection with low false positive rates. ML classifier running at 2ms inference. Pre-filter blocks obvious attacks 1,550x faster than LLM-only.
Deception engine
Drawing on deep cyber deception expertise, we built three deception modes (honeypot, tarpit, redirect) that turn detected attacks into intelligence-gathering operations.
Open source launch
Oubliette Shield published on PyPI under Apache 2.0. 12 LLM providers, 9 SDK integrations, 280+ automated tests. Because every AI system deserves protection.
Our Expertise
Purple team expertise across the cybersecurity landscape — bridging offense, defense, deception, and intelligence.
Purple Teaming
Bridging offense and defense. We combine adversarial testing with detection engineering to build security that improves with every attack.
Cyber Deception
Honeypots, honey tokens, decoy systems, and attacker misdirection. Our core philosophy: make the attacker think they succeeded.
Threat Hunting
Proactive detection of adversary activity across networks and AI systems. Finding the threats that automated tools miss.
Threat Intelligence
IOC extraction, STIX 2.1 export, MITRE ATLAS mapping, and feed ingestion. Turning raw attack data into actionable intelligence.
Red Teaming
Adversarial testing of AI systems and traditional infrastructure. Understanding attacker techniques firsthand to build better defenses.
Founded by a Veteran
Oubliette Security was founded by a veteran with a background in purple teaming — cyber deception, threat hunting, and threat intelligence. Our approach to AI security comes directly from this experience: the best defense doesn't just detect threats, it deceives and studies them.
We're currently pursuing SDVOSB (Service-Disabled Veteran-Owned Small Business) certification to better serve federal and defense customers.
We believe the best defense is one that turns the attacker's own techniques against them.
Get in TouchOpen Source First
Oubliette Shield is open source under Apache 2.0. We believe security tools should be accessible to everyone, not just enterprises with six-figure budgets.