Research Project: This is a free AI research project. No warranties, SLAs, or company associations. Learn more

Managed Detection & Response

Detection, Response, and Learning in 10-Second Cycles

65 signal types, 30 attack chain patterns, and a rehearsal engine that responds in under 500ms. Every outcome makes the system smarter.

65+
Signal Types
<500ms
Rehearsed Response
10s
OODA Cycle
30
Attack Patterns

The OODA Loop

Observe, Orient, Decide, Act — every 10 seconds, for every signal

OBSERVE

Signal Collection

Signals from 42+ connectors, edge agents, and SIEM event lake — normalized and enriched in real-time.

  • Every 10 seconds, the reasoning loop processes queued signals with MITRE ATT&CK mapping and risk scoring
ORIENT

Brain Analysis

EvolutionaryBrain recalls similar past decisions via pgvector, applies Bayesian confidence from prior outcomes.

  • The brain embeds each signal as a 1024-dimensional vector and retrieves the 5 most similar historical decisions
DECIDE

Risk-Tiered Decision

Risk-tiered decision framework: LOW auto-executes, MEDIUM needs 85%+ confidence, HIGH requires human approval.

  • Every decision includes confidence score, risk assessment, alternative actions considered, and rollback plan
ACT

Scoped Remediation

Scoped ChangeSets with blast-radius assessment, policy verification, DRY_RUN preview, and tested rollback.

  • 6 production templates: ISOLATE_ENDPOINT, DISABLE_ACCOUNT, QUARANTINE_EMAIL, CONTAIN_LATERAL, ROLLBACK_POLICY, RANSOMWARE_PREEMPT
Rehearsal Engine

Pre-Compiled Responses in Under 500ms

The rehearsal engine pre-computes responses for known attack scenarios against each tenant's security profile. When a matching signal arrives, response is instant.

  • Employee in HIBP breach, no MFA — pre-compiled: force reset + enroll MFA + 24h monitoring
  • RDP exposed to internet — pre-compiled: OPA policy push + evidence + notify insurance
  • BYOD jailbreak with corporate data — pre-compiled: selective wipe + conditional access block
  • Admin credential in dark web — pre-compiled: disable account + force re-enrollment + incident ticket

Attack Chain Correlation

Multi-stage detection links individual signals into coordinated attack patterns

Credential Stuffing

Breach Detected then Login Attempts then Lateral Movement

Initial Accessto Lateral Movement

Ransomware Precursor

RDP Exposed then Brute Force then Privilege Escalation

Reconto Impact

BYOD Data Exfiltration

Non-Compliant Device then Corporate Access then Unusual Download

Initial Accessto Exfiltration

Supply Chain Identity

Vendor Breach then Phishing from Trusted Sender then Credential Harvest

Reconto Credential Access

Insider Threat

Privilege Escalation then Sensitive Data Access then Exfiltration Pattern

Priv Escto Exfiltration

Identity to Lateral

Breach then Same Password on Corp then Login then Lateral Movement

Initial Accessto Lateral Movement

MDR FAQ

MDR That Actually Responds

Detection without response is just a dashboard. BrainstormMSP detects, responds, proves, and learns — every 10 seconds.