Active Defense, Threat Telemetry & AI SOC

Yosef Moshe

Security Engineering Portfolio: Firewalls, Identity, SIEM & Autonomous AI Triage

pfSense Wazuh SIEM Suricata Local LLM (Ollama) AbuseIPDB API Samba AD-DC CDB Hash Tables Active Response MITRE ATT&CK Python
Perimeter & Detection Engineering

Stateful Telemetry & Threat Intelligence Loop

A segmented enterprise lab combining pfSense stateful firewall inspection, Samba Active Directory identity telemetry, and an automated threat intelligence loop.

  • BSD filterlog Parsing (OS_Regex) Engineered custom parent/child decoders in Wazuh, resolving complex RFC-5424 header mismatches to cleanly extract IPs and ports across TCP/UDP/ICMP.
  • Stateful Timeframe Correlation Configured 60-second correlation windows to filter background noise, triggering threat intel API queries only on persistent adversary scans.
  • Dynamic CDB Compilation Automated Python integration that queries AbuseIPDB, appends confirmed threats (>20% score), and instantly compiles binary .cdb lists in memory.
  • Bi-Directional Feed Blocking Deployed Level 12 SIEM rules to instantly isolate inbound attacks from verified actors or internal endpoints attempting outbound C2 communication.
flowchart TD Kali["pitakli
(Kali Attacker)"] subgraph Perimeter ["Perimeter & Identity"] pfSense["pitasense (pfSense)
Table: wazuh_blacklist"] DC["dcpita (Samba AD-DC)"] end subgraph Detection ["SIEM & Threat Intel"] WAZ["pitawazuh
Wazuh Manager"] Abuse["AbuseIPDB API
Confidence >20%"] CDB[("malicious-ip
CDB Binary")] end Kali -.->|"Drops & Scans"| pfSense Kali -.->|"Kerberos & SMB"| DC pfSense -->|"BSD Syslog (514)"| WAZ DC -.->|"Agent Logs"| WAZ WAZ -->|"Session >60s"| Abuse Abuse -->|"Auto-Compile"| CDB CDB -->|"pfctl Table Sync"| pfSense classDef srv fill:#1e293b,stroke:#38bdf8,stroke-width:1.5px,color:#fff; classDef sec fill:#0f172a,stroke:#f43f5e,stroke-width:1.5px,color:#fff; class Kali,pfSense sec; class DC,WAZ,Abuse,CDB srv;

Live Execution Telemetry

AI-Driven SOC & SOAR

Autonomous Local AI SOC Analyst

A private, on-premises AI incident response pipeline utilizing a local LLM to digest live SIEM alerts, map MITRE ATT&CK tactics, generate professional incident reports, and enforce zero-touch firewall isolation.

  • Private Local AI Triage Dispatches SIEM alerts to a local Ollama instance (Llama 3), classifying alerts without exposing sensitive telemetry to cloud APIs.
  • MITRE ATT&CK Mapping Autonomously identifies IOAs/IOCs and contextualizes adversary behavior (e.g., T1110 Brute Force, T1046 Network Service Discovery).
  • Zero-Touch Active Response & Whitelisting Executes kernel packet filtering directly on pfSense via passwordless SSH, protected by fail-safe logic to prevent core infrastructure isolation.
  • SIEM Loopback & Reporting Generates Markdown incident reports locally while streaming AI verdicts back into the Wazuh Dashboard as native alerts via custom JSON ingestion.
flowchart TD subgraph Input ["Alert Pipeline"] Log["alerts.json
(Severity 6+)"] end subgraph AI ["Local AI Orchestration"] Agent["soc_ai_agent.py
(Daemon)"] LLM["Ollama Llama 3
(Local Inference)"] end subgraph Output ["Enforcement & Reporting"] FW["pfSense
Block Table"] Rep["Markdown
Report"] Dash["Wazuh
Dashboard"] end Log -->|"Tail Events"| Agent Agent <-->|"JSON Schema"| LLM Agent -->|"Ed25519 SSH"| FW Agent -->|"Save to Disk"| Rep Agent -->|"Log Loopback"| Dash classDef core fill:#0f172a,stroke:#38bdf8,stroke-width:1.5px,color:#fff; classDef ai fill:#3b0764,stroke:#d946ef,stroke-width:1.5px,color:#fff; classDef action fill:#064e3b,stroke:#10b981,stroke-width:1.5px,color:#fff; class Log core; class Agent,LLM ai; class FW,Rep,Dash action;

Autonomous AI Execution Proofs

AI-Driven SOC Telemetry & Rule Synthesis

AI & SecOps

Automated SOC pipeline using LLM agents to ingest raw telemetry, isolate anomalous attack vectors, and automatically generate validated Wazuh and Suricata detection rules.

Automated Event Ingestion: Normalizes unstructured syslog-ng, firewall drops, and Wazuh alerts into LLM context windows to extract actionable indicators of compromise (IoCs).
Dynamic Rule Synthesis: Translates verified attack chains directly into tuned XML detection rules for Wazuh agents and Suricata signatures, mitigating rule fatigue.
Incident Triage Summaries: Generates executive and Tier-2 incident briefs mapped to MITRE ATT&CK tactics within seconds of alert escalation.
Python Wazuh Rules XML LLM Agent / API