Red Specter Security Research Ltd

Red Specter Security Research Ltd designs and operates purpose-built frameworks for the complete attack surface of agentic AI systems. Spanning offensive research, runtime defence, forensic incident reconstruction, and autonomous red-blue co-evolution.

What We Build

NIGHTFALL

Offensive AI security framework. 312 purpose-built tools across 246 attack layers and 37 kill chain phases. The complete agentic AI attack surface.

AI SHIELD

Defensive runtime platform. 239 modules across 17 industry verticals. Purpose-built runtime protection paired to every NIGHTFALL attack surface.

RED SCORE

AI security posture scoring. Quantified, evidence-based risk assessment across the full agentic attack surface.

BLACK BOX

Forensic AI incident reconstruction. Cryptographically-linked capture of every input, output, tool invocation, and reasoning checkpoint — before, during, and after an incident.

SPECTER BATTLE LAB

Autonomous attack-defence co-evolution platform. NIGHTFALL attacks. AI Shield defends. Security Half-Life empirically measured at T½ = 13 cycles.

The SPECTER Constellation

SPECTER AGENT RED

Offensive intelligence. Sequences attacks across 301 NIGHTFALL tools and 22 attack surfaces. Ed25519 + ML-DSA-65 + RFC 3161 evidence chain. SAR-{hex12} session IDs.

SPECTER AGENT BLUE

Defensive coverage mapping. Maps 227 AI Shield modules against 22 attack surfaces. Cross-references RED findings. Flags untested and undefended surfaces. SAB-{hex12} session IDs.

SPECTER AGENT WHITE

Compliance and regulatory mapping. Takes RED and BLUE outputs. Maps findings to 11 regulatory frameworks and 37 obligations. SAW-{hex12} session IDs.

Attack. Defend. Comply.

Coming soon: SPECTER AGENT BLACK (threat intelligence), SPECTER AGENT GREY (forensics), SPECTER AGENT GOLD (vendor assessment), SPECTER AGENT PURPLE (continuous monitoring), SPECTER AGENT ORANGE (training and simulation).

Engagements

Red Specter runs the full cycle. NIGHTFALL attacks your infrastructure in a controlled lab environment. AI Shield tells you exactly what it found and how to close it. Every engagement produces empirical findings — not a generic report.

Fixed-Scope Assessment

A defined engagement against your agentic AI stack. NIGHTFALL sequences attacks across your declared surface. RED SCORE quantifies your posture. You receive a signed findings report with AI Shield remediation recommendations for every finding. Delivered in two weeks.

Full Engagement

Red team plus defensive build. NIGHTFALL attacks your stack. AI Shield modules are purpose-built from the empirical findings — tuned to your infrastructure, not generic coverage. You finish the engagement with working defensive runtime protection, not a PDF. Timeline agreed at scoping.

Continuous Monitoring

Ongoing red-blue co-evolution against your stack. SPECTER BATTLE LAB runs continuous attack-defence cycles. AI Shield updates as NIGHTFALL finds new paths. Security Half-Life tracked and reported. Retainer-based.

Discuss an Engagement →

Research

114 Published Papers — Peer-reviewed research across offensive AI, defensive runtime protection, and AI forensics.

View on Zenodo →

RS-2026-081: The SPECTER Constellation. doi:10.5281/zenodo.22962665

RS-2026-068: The NHI Blind Spot. doi:10.5281/zenodo.22714844

RS-2026-067: The DEADMAN Problem. doi:10.5281/zenodo.22709674

RS-2026-066: SPECTER TORPEDO. doi:10.5281/zenodo.22709618