Hexward Labs

Self-hosted security tooling

Tools I needed, so I built them.

Most of what follows started as a script I wrote because the commercial option cost more than the problem, or ran in someone else's cloud, or only converted in one direction. Each one is a single Go binary with SQLite storage. Nothing phones home.

All source on GitHub Firewall migration after Expedition

The line

RuleForge

Multivendor firewall migration. Converts between any two of Cisco ASA, FTD, PAN-OS, FortiGate and Check Point — twenty directions, through a vendor-neutral model.

RuleHawk

Firewall config auditing. Finds rules shadowed by anything above them, allows far broader than the ticket that created them, and drift against a baseline.

Decoy

Canary tokens and honeypots. Bait that no legitimate user has a reason to touch — so when one trips, it is almost never a false alarm.

Attack Surface Monitor

Enumerates your exposure the way a stranger would, on a schedule. Catches the staging host and the dangling DNS record nobody remembers.

Patchlight

CVE prioritisation. Ranks your actual inventory by CISA KEV and EPSS, not by CVSS alone. Patch what is being exploited, not what scores loudest.

CertLight

TLS monitoring past the expiry date — hostname mismatch, broken chains, weak keys, and hosts still accepting obsolete protocols behind a valid certificate.

Loglight

SIEM-lite. Ingests syslog, files, journald, Docker and Windows events, and correlates a scan followed by brute force followed by a successful login into one incident.

DmarcWatch

DMARC monitoring. Reads the aggregate reports nobody opens — who sends as your domain, who is spoofing it, and when it is safe to move to p=reject.

TopoLight

Network monitoring with a live LLDP topology map — and when a switch dies, everything behind it folds under the one root cause instead of paging you thirty times. 0.4 adds NetFlow/IPFIX/sFlow, MAC/ARP endpoints, BGP/OSPF/STP state, wireless (UniFi, Meraki, Cisco WLC, Aruba), SD-WAN health, SSH config backup with diff, probes, reports, a 2–5 node cluster and a gNMI beta. One static binary.

TenantWatch

Read-only Microsoft 365 and Google Workspace posture — accounts without MFA, over-permissioned OAuth apps, mailboxes forwarding outside the org. Since 0.2 it also reads the sign-in log: legacy authentication that actually succeeded, an admin who signed in with a password alone, a password spray that landed.

Posture Report

Reads every other tool's findings, read-only, and folds them into one posture score with an executive page and a technical page. It runs no scans of its own and changes nothing — it answers "how are we doing, and what do we fix first" on paper an auditor accepts.

AuditLight

Security assessment that never exploits — no brute force, no denial of service, no fuzzing — so it is safe to point at production. Three reports: the assessment you hand over, the process report naming every check attempted and every target skipped with its reason, and the change report that shows whether the fixes held.

Every tool can emit its findings as RFC 3164 syslog, which is how they feed each other — or feed whatever collector you already run. There is nothing Hexward-specific about the format.

Written up in detail

How these tools are built

Not marketing claims — constraints I hold myself to, so you can check whether I kept them.

Your data does not leave your network

Every tool runs on your infrastructure as one binary. There is no telemetry, no analytics, no usage reporting. Licence validation is offline Ed25519 cryptography: a key is verified locally and the software never needs to reach any server of mine. An expired key does not disarm anything — it returns the tool to free limits.

The free edition is genuinely usable

Free editions are Apache-2.0 and complete enough to solve a real problem, not a demo with the working parts removed. Where a limit exists it is stated on the repository page before you install anything.

Nothing is silently dropped

Where a tool cannot handle something, it says so and shows you the source line. A migration that quietly discards a rule, or a detector that quietly fails to parse a timestamp, is worse than no tool at all — because it leaves you confident and wrong.

Limits are published next to the features

Every repository has an “honest limits” section covering what the tool does not do. Deception is not prevention. A vulnerability ranking is not a patch. An attack surface scan sees what is reachable, not what is safe.

Who wrote this

Nizar Tuanku — Cybersecurity. Mostly enterprise networks and Cisco security. I build these in the evenings because I keep meeting the same problems and wanting the same tools.

Source, issues and pull requests: github.com/nizartuanku. Paid editions, for the cases where the free limits are not enough, are on Whop.