CCNP Enterprise, Architecture and Design · CySA+ · Security+ · based in New York · open to roles in Germany, across Europe and in the US
Fourteen years securing enterprise and industrial networks. I scope, design, build, automate and document, and I am strongest where security, network and automation meet. Everything public here is rewritten from scratch and carries no customer data.
Hiring? What I do · Recent field work · Field notes · Contact | Hands-on? IT to OT · Battle drill · Which tool, when · Try one in sixty seconds · Playbooks
| Role | What it looked like |
|---|---|
| OT/ICS security architect | Security architecture for the operational networks of a large US transit agency: identity, segmentation and detection across train control, SCADA, fire alarm, elevators, fare collection and CCTV. Security groups mapped to IEC 62443 zones under the TSA rail security directives. |
| Identity and Zero Trust lead | Cisco ISE 3.x design, upgrades and recovery, 802.1X EAP-TLS, closed-mode MAB, TrustSec, PKI. Took over a biotech campus with broken authentication and stabilized it: 915 endpoints on certificate-based EAP-TLS across eight virtual networks. |
| Visibility and detection | Network TAPs on train control feeding two OT detection platforms, cut in during one overnight window with no service impact. Traced 85 percent of daily alarming hosts in a flow analytics platform to ten rules, then worked them down host by host instead of switching rules off. |
| Automation at scale | Python AAA hardening pushed to 6,000+ switches. Passive asset discovery that found thousands of OT devices missing from every inventory. The tools below are the public versions of this work. |
| Networks and voice | SD-Access and Catalyst Center, Catalyst 9800 wireless, Nexus vPC, Secure Firewall. Moved a state agency's 17 sites to SD-WAN, dual stack, with BGP to Azure. Leading a county's phone system refresh: Unified Communications Manager and Unity 11.5 to 15, nine PRI circuits moving to SIP on two CUBE gateways. BGP over AWS Direct Connect for a genomics lab. |
| Planning and delivery | Requirements turned into controls and evidence. Level of effort, statements of work and bills of material checked before the order. Phased plans with a decision point and a rollback at every step, such as a data center switch replacement under a live vSAN stretch cluster, moved three server links at a time so no server ever lost its path. |
Platforms I work in: Cisco ISE, Catalyst Center, SD-Access, Catalyst 9800, Meraki, Nexus, Secure Firewall, Secure Network Analytics, Gigamon, Nozomi, Darktrace, Python with Netmiko, PowerShell.
The Purdue model from the enterprise down to the field, and what I do at each level. Clickable version.
%%{init: {"flowchart": {"rankSpacing": 30, "nodeSpacing": 20}}}%%
flowchart LR
L4["<b>Levels 4 and 5: Enterprise IT</b><br/>ISE identity and Zero Trust<br/>SD-Access, Catalyst Center<br/>firewalls, cloud links"]
L35["<b>Level 3.5: OT DMZ</b><br/>jump hosts, patch relay<br/>historian replica<br/>remote access with MFA"]
L3["<b>Level 3: Operations</b><br/>control center servers<br/>TrustSec security groups<br/>NDR sensors"]
L2["<b>Level 2: Supervisory</b><br/>HMIs, station systems<br/>closed-mode MAB<br/>for devices without 802.1X"]
L1["<b>Levels 1 and 0: Control and field</b><br/>train control, PLCs, signals<br/>agentless, passive TAPs<br/>data diode out"]
L4 -->|conduit:<br/>firewall, jump host| L35
L35 -->|conduit:<br/>firewall, monitoring| L3
L3 <-->|security groups,<br/>MAB| L2
L2 ~~~ L1
L1 -.->|one way out,<br/>data diode| L2
style L4 fill:#1f6feb,color:#fff,stroke:none
style L35 fill:#6e40c9,color:#fff,stroke:none
style L3 fill:#1a7f37,color:#fff,stroke:none
style L2 fill:#1a7f37,color:#fff,stroke:none
style L1 fill:#cf222e,color:#fff,stroke:none
| Level | What I do there | Tools |
|---|---|---|
| 4 and 5, enterprise IT | ISE identity and Zero Trust, SD-Access, Catalyst Center, firewalls, cloud links | ise-endpoint-explain, ise-health-audit, cisco-interface-health |
| 3.5, OT DMZ | Jump hosts, patch relay, historian replica, vendor access with MFA; every conduit checked against real traffic | ot-zone-conduit-check |
| 3, operations | Security groups across train control, SCADA, fire alarm, elevators, fare collection and CCTV, mapped to IEC 62443; NDR sensors | ndr-coverage-map, cisco-config-drift |
| 2, supervisory | Closed-mode MAB for devices that cannot run 802.1X, AAA hardening, asset discovery | ot-asset-inventory, aaa-lockout-check, ios-hardening-audit |
| 1 and 0, control and field | Safety systems stay agentless; passive TAPs, data diode out | ndr-coverage-map |
The Purdue chart above shows where the work happens. This shows how: the six steps every change goes through, from the order to the objective held. Each step ends with a go/no-go check, and a no-go sends the work back along the red line. Open the interactive version to walk it step by step with two worked examples.
| Step | What happens | Go/no-go check | Tools and playbooks |
|---|---|---|---|
| 1. Orders | The directive, audit finding or incident becomes a control with a scope and a pass test. | None, straight to recon. | Two sources and a proof |
| 2. Recon | Endpoints per switch port from offline show output, every running-config baselined and audited, gaps listed by zone. Nothing is sent to the OT devices. | Ground known? No-go: keep recon. | ot-asset-inventory, ios-hardening-audit, cisco-config-drift |
| 3. Plan | Zones and conduits, 802.1X or closed-mode MAB, TrustSec groups and contracts, checked against observed flows until nothing legitimate breaks. | Plan survives real traffic? No-go: re-plan. | ot-zone-conduit-check, Closed-mode MAB, one site at a time |
| 4. Rehearse | Written procedure with pre-checks, commands, verification and rollback; lockout check; ISE health before the window; a script that stops at the first failed check. | Lockout and rollback clear? No-go: hold, fix, rehearse again. | aaa-lockout-check, ise-health-audit, Production change by script, The script contract |
| 5. Execute | One site, or one batch of ports or switches, at a time; each proven before the next moves. | Every check passed? No-go: roll back the batch, back to step 4. | ise-endpoint-explain, cisco-interface-health, cisco-config-drift |
| 6. Hold the ground | Drift against the approved state, ISE health and certificate expiry, detection tuned host by host. | Anything new found? Yes: new orders, back to step 1. No: objective held. | cisco-config-drift, ise-health-audit, ndr-coverage-map, NDR alarm noise reduction loop |
Anonymized.
- Moved 25 security groups and 9 contracts from ISE to Catalyst Center Group-Based Access Control with zero failures.
- Upgraded a Catalyst Center cluster through 2.3.7 and on to 3.2.3 in one overnight window.
- Proved that stalled uploads blamed on the network came from a saturated 10G cloud circuit, by matching controller interface history against the carrier's graphs. A larger circuit was approved the same day.
- Reproduced a controller provisioning fault that wiped the WLAN mappings on a Catalyst 9800, under a five-second watch with automatic restore.
- Fixed an EAP-TLS incident (ISE failure 22047) by adding UPN and URI subject alternative names to the MDM's SCEP user certificate profile.
- Built and run the knowledge base every engagement works from: one Cloudflare Worker, a sealed container per knowledge base, sign in through an allow list, every write versioned, a secret guard on writes, scheduled backups, and a deploy pipeline that smoke tests and rolls back.
| The situation | Run | What you get back |
|---|---|---|
| "What is on this OT network?" and nobody may scan it | ot-asset-inventory | Endpoints per switch port from ARP, MAC, interface and CDP/LLDP output, hidden switches and MACs with no IP flagged, and a diff of new, gone and moved devices between runs |
| Moving switch logins to ISE TACACS+ tomorrow | aaa-lockout-check | Every AAA shape in the config that can lock you out: no working fallback, console authorization, undefined lists, a dead source interface, a management VRF mismatch |
| "Why did this laptop land in the wrong VLAN?" | ise-endpoint-explain | The live session plus an authentication timeline: switch and port, authorization profile, SGT, matched rule, and every failure reason in plain English |
| Change window tomorrow. Is ISE healthy? | ise-health-audit | Nodes down, missing admin HA, system and trusted certificates expired or expiring within 30 and 60 days |
| "Did anyone touch the switches since the last approved state?" | cisco-config-drift | A baseline of every running-config and a clean diff of anything changed outside change control |
| "The network is slow" or a port keeps flapping | cisco-interface-health | Ports ranked by urgency: err-disabled, CRC, duplex mismatch, drops, resets, each with the likely cause and the next step |
| New site handover, or audit prep | ios-hardening-audit | Every hardening gap per device with severity, the evidence line and the exact fix |
| A new or inherited Windows machine | win-security-posture-audit | TPM, Secure Boot, Defender, BitLocker, firewall, startup items, drivers and hardware health in one report, plus a fix plan it does not apply |
Four of the tools run entirely offline on text you already have. Three of them:
# Paste "show interfaces" output from one switch or fifty; get the ports that need attention, ranked
pip install git+https://github.com/a-ayrapetov/cisco-interface-health
ifhealth show_interfaces.txt
# Point it at a folder of running-configs; get every hardening gap with the fix
pip install git+https://github.com/a-ayrapetov/ios-hardening-audit
ios-harden configs/*.cfg --csv findings.csv
# Same folder of configs; find what would lock you out of a TACACS+ change before you make it
pip install git+https://github.com/a-ayrapetov/aaa-lockout-check
aaa-lockout-check "configs/*.cfg"How the work is run, written from the same engagements with no customer in them.
| Playbook | When to read it |
|---|---|
| Closed-mode MAB, one site at a time | You are closing access ports for devices that cannot do 802.1X |
| NDR alarm noise reduction loop | Your flow analytics platform alarms all day and nobody looks any more |
| Production change by script | A change has to reach a production device and the way back must be ready before the first line goes in |
| The script contract | You write automation that someone else runs on a host you cannot reach |
| Two sources and a proof | Someone asks "how much", "where" or "who" about a network and the answer will drive a decision |
The checks in the change-by-script and script-contract playbooks are also code: netops_gates_and_proofs.py, standard library only, 35 tests.
Build status
Rules every tool follows
Read-only by default. Show commands and GET calls. Nothing is pushed to a device unless the README says so in its first line.
No stored credentials. Prompted at runtime or read from the environment, never written to disk, never logged.
No real networks in the repo. Sample data uses RFC 5737 documentation addresses only.
Gated CI. Every push runs a secret scan and fails on any private RFC 1918 address. The same gate is what keeps customer data out of this account.
Evidence over opinion. Each tool prints the lines that prove its own result, so the output can go straight into a change record.
New tools start from repo-template and land here once they pass review.
Longer write-ups of the same work on a-ayrapetov.github.io:
- Inventory an OT network without sending it a single packet
- Moving switches to ISE TACACS+ without locking yourself out
- Closing access ports on industrial switches, one port at a time
- Cutting NDR alarm noise without going blind
- An AI assistant on a production network that never gets a session
- When TrustSec looks like a routing problem
Open to architect roles in network and OT security, in Germany, across Europe and in the US, and to consulting work. Email arkadiy@ayrapetov.co · LinkedIn · field notes on network and OT security at a-ayrapetov.github.io

