Behavior profiler
PROBIZENT
Per-process baselining flags signature-clean drift.
Process, file, and network behavior features establish a local baseline and flag drift. Behavior is supporting evidence in the composite decision pipeline; a single anomaly does not authorize destructive remediation.
Allowlist-based execution. Unknowns queued for approval.
Unknown binaries can be held for review after exact hashing and identity checks. Protected operating-system targets fail closed, while publisher, reputation, ML, and behavioral evidence determine whether an unfamiliar file is allowed, blocked, or queued.
Fileless attack detector
PROBIZENT
PowerShell abuse, WMI, COM hijack, reflective injection.
Inspects PowerShell, WMI, script, LOLBin, and injection indicators and uses AMSI when available. The resulting evidence feeds the same composite policy and does not claim complete coverage of fileless techniques.
Ransomware correlator
PROBIZENT
Entropy, decoy, and behavior evidence drive exact containment.
When independent ransomware signals converge on the same verified process identity, Defendrix can suspend it and create a durable Pending Review record. Resume or terminate actions revalidate PID, creation time, image path, and file identity before acting; Defendrix does not advertise nonexistent user-file rollback.
Zero-trust execution
PROBIZENT
Unknown binaries require corroborated evidence before allow.
Zero-trust execution correlates exact hash reputation, publisher evidence, local ML scoring, and bounded behavioral evidence. Invalid or incomplete evidence never becomes an allow verdict; low-confidence signals require behavioral corroboration.
File provenance chain
PROBIZENT
Per-file Merkle chain catches supply-chain substitutions.
Observed writes to executables and scripts add authenticated provenance records containing the prior hash, new hash, and writer identity. A mismatch raises tamper evidence for the decision pipeline instead of trusting a filename, folder, or certificate alone.
Adversarial LLM detection
PROBIZENT
Detects AI-generated phishing + malicious scripts.
A statistical classifier scores text and scripts for adversarial automation indicators. It is deliberately a soft signal: it contributes to composite severity but never authorizes quarantine on its own.