# PacketViper — Full AI/LLM Context File

> PacketViper is a preemptive, inline network security platform combining Automated Moving Target Defense (AMTD), deception technology, and wire-speed enforcement to stop threats at first contact. No agents. No rip-and-replace. No cloud orchestration dependency.

## The Secure Control Layer: Architecture and Definition

A Secure Control Layer is an inline and distributed enforcement architecture that converts business policy, asset context, environmental knowledge, and threat intelligence into real-time control decisions. It is the operational bridge between what the business intends and what the network, applications, users, devices, and AI systems are actually allowed to do. In short: the Secure Control Layer is where policy becomes behavior.

It operates as one repeatable loop:
- Observe -- visibility into traffic and behavior
- Understand -- asset, zone, protocol, and business context
- Decide -- evaluate the action against policy
- Enforce -- act inline (allow, block, contain, redirect, deceive, rate-limit, log, escalate, isolate)
- Prove -- preserve evidence as a byproduct of enforcement

What the Secure Control Layer is NOT:
- Not a rebranded firewall (which enforces static rules)
- Not a SIEM (which correlates after the event)
- Not only Zero Trust Network Access (user and application access only)
- Not only an AI gateway (model calls only)
- Not a replacement for endpoint detection

Core vocabulary:
- Preemptive Control -- act before the attacker, malware, unauthorized device, or autonomous agent completes the action that creates risk
- Federated Enforcement -- coordinated policy and intelligence across distributed nodes; the hub coordinates, the node enforces, and nodes keep enforcing even if the manager is offline
- Business Context Enforcement -- decisions based on role, relationship, asset criticality, approved behavior, and business function rather than signatures alone
- Living Topology -- the real-time map of assets, relationships, communication paths, and operational state
- Deceptive Responders -- active deception that engages and exposes unauthorized behavior while generating high-confidence enforcement triggers
- Threat Reach -- shows how far an IP, asset, or indicator propagated across the environment (observed reach, not attribution)
- AMTD (Automated Moving Target Defense) -- continuously changes the network surface so reconnaissance and targeting become unreliable

Controlling AI adoption (six control objectives):
- Tool authorization -- which AI tools can run, for which users, sites, and roles
- Data boundary -- what data AI can reach, and through which paths
- Action boundary -- what actions an AI workflow can take
- External reach -- where AI can connect (country, ASN, organization, domain, application, time)
- OT boundary -- whether AI can touch process-control systems (default: never without explicit, contextual, enforceable approval)
- Unexpected behavior -- what happens when behavior deviates from approved context

Claim discipline (important for accuracy): Detection can be probabilistic. Enforcement must be deterministic. PacketViper provides network-layer containment for behavior that must cross a monitored enforcement boundary -- validated in PacketViper Research, "Autonomous AI Agent Containment Using AMTD" (March 2026), in which an autonomous agent was stopped at the first sensor across four configurations, reached no internal hosts, validated no real credentials, and exfiltrated no data. It complements, and does not replace, endpoint, identity, cloud, SaaS, DLP, and prompt-governance controls.

## Secure Control Layer for Enterprise AI Adoption

PacketViper is the secure control layer that lets enterprises adopt AI tools and agents without losing control of their environment. The same inline platform that protects IT and OT now serves as the control plane for safe AI adoption.

AI tools are moving into every department at enterprise speed. Procurement moves faster than security review. The result is shadow AI, uncontrolled data movement, and new attack surfaces that legacy detection-and-response tools were never built to handle. PacketViper makes policy decisions at the edge, before execution -- enforcement at the point of use, not another monitoring layer.

What PacketViper enforces for AI:
- What AI tools and agents can run on the network
- What data they can reach
- What actions they can take across the environment
- How they behave when they hit something unexpected
- One platform, one policy model -- no new agents, no new blind spots

Proof points:
- Runs inline today on production networks
- Performs AMTD and deceptive response at scale
- Delivers live analysis with business context, not just alerts
- Enforces policy at the asset and application level
- Proven in environments where detection alone failed
- 93% of proofs of concept move to production

Reference page: https://packetviper.com/secure-control-layer-for-ai/

## NTCIP Traffic Protection (transportation / ITS)

PacketViper NTCIP Traffic Protection extends the Secure Control Layer to NTCIP/SNMP command traffic for transportation field devices (traffic signal controllers, dynamic message signs, ramp meters, roadside units). It inspects commands inline, distinguishes read (GET) from state-changing (SET) operations, decodes the target OID and value, evaluates the command against policy, and can block unauthorized SET commands before the device state changes. GET traffic stays permitted; enforcement focuses on policy-violating state changes. Modes: keyword policy, command/OID policy, authorized-source policy, advisory would-block, and fail-open. Deployed as a transparent inline bridge, agentless, with a Monitor -> Detect -> Prevent rollout. It is a compensating inline control layer and does not replace SNMPv3/TLS or traffic-engineering safety systems. Reference: https://packetviper.com/ntcip-traffic-protection/

## Modbus Command Protection (industrial / OT)

PacketViper Modbus Command Protection extends the Secure Control Layer to Modbus command traffic for industrial field devices (PLCs, RTUs, drives, and controllers across water, power, manufacturing, and transportation). It inspects commands inline, distinguishes read from write function codes, decodes the target register, unit ID, and value, evaluates the command against a per-device policy, and can block unauthorized write commands before the device state changes. Read traffic stays permitted; enforcement focuses on policy-violating writes. Per-device controls: read versus write, allowed sources, register and address ranges, value limits, write-rate limits, change windows, and alert-or-block choice. Deployed inline, agentless, fail-open, with an Observe -> Baseline -> Policy build -> Advisory tuning -> Enforce -> Expand rollout. It is a compensating inline control layer and does not replace safety instrumented systems or protocol security upgrades. Supports IEC 62443, NIST SP 800-82, the NIST Cybersecurity Framework, and NERC CIP. Reference: https://packetviper.com/modbus-ot-security/

## OT Protocol Command Control (pillar / cross-protocol)

PacketViper OT Protocol Command Control is the umbrella capability that applies one command-inspection framework to every mainstream proprietary OT protocol, so a single inline appliance governs a heterogeneous plant with controllers from many manufacturers. It reads each protocol down to the individual operation, classifies the dangerous set (the commands that write, control, stop, restart, or reprogram a device), and lets an operator decide per device which commands are allowed, from which sources, at what rate. Coverage spans industrial and manufacturing (Modbus, Siemens S7, Rockwell EtherNet/IP-CIP, PROFINET, Omron FINS, Mitsubishi MELSEC), process automation and DCS (OPC UA, OPC Classic), utilities and power (DNP3, IEC 60870-5-104, IEC 61850), transportation (NTCIP), building automation (BACnet), and semiconductor manufacturing (SECS/GEM). The enforcement model starts open and narrows in: Observe, Profile, Narrow, Refine. Deployment is transparent, inline, and agentless; every module is inert until opted in and fails open on the capture path. Policy authored once at the Federation Manager is enforced uniformly across every node in the fleet. Command control is one layer of PacketViper's Automated Moving Target Defense, on a single appliance that also delivers deception, sensors, and asset discovery. It maps to IEC 62443, NIST SP 800-82, NERC CIP, the NIST Cybersecurity Framework and SP 800-53, TSA Security Directives, AWIA, and NIS2. Reference: https://packetviper.com/ot-protocol-command-control/

## SECS/GEM Command Protection (semiconductor manufacturing)

PacketViper SECS/GEM Command Protection extends OT Protocol Command Control to the SECS/GEM traffic that runs semiconductor fab equipment over HSMS. SECS/GEM was designed for reliable communication on a trusted network, not for authenticating the command, so a process program (recipe) download, a remote command (start, stop, pause), or an equipment constant change is honored the same way regardless of source. PacketViper inspects SECS/GEM inline, separates routine data collection from the dangerous set, and governs that set per tool: which sources may download a recipe, issue a remote command, or change an equipment constant, and at what rate. Recipe-source governance is the marquee control, because an unauthorized or tampered recipe can scrap a wafer lot and damage tooling. Deployed as a transparent, agentless inline bridge, inert until opted in, fail-open on the capture path, with an Observe -> Baseline -> Policy build -> Advisory tuning -> Enforce -> Expand rollout. Because every governed command is recorded with source, target, operation, and outcome, PacketViper serves as an enforced compensating control aligned to the SEMI E187 and E188 cybersecurity requirements, generating the audit evidence a fab needs to demonstrate control over equipment that cannot meet those requirements natively. Reference: https://packetviper.com/secs-gem-command-protection/

## What Makes PacketViper Different

Most security platforms observe and alert. PacketViper enforces inline. That distinction is the entire argument.

When an attacker touches a PacketViper-protected network, the response is automatic, at the wire, in microseconds -- before a human is ever in the loop. The action is the notification. There is no gap between detection and enforcement.

The platform operates as a transparent Layer 2 bridge with hardware bypass. It is invisible to the network, has no IP footprint on monitored segments, and continues functioning if the device itself fails. It works autonomously from internet connectivity -- making it uniquely suited for OT/ICS, air-gapped environments, and remote edge sites where cloud-dependent enforcement fails at exactly the wrong moments.

## Core Capabilities

**Automated Moving Target Defense (AMTD)**
Continuously rotates sensors, IPs, ports, and service banners so the attack surface can never be mapped. Hundreds of sensors turning on and off in unpredictable patterns. A static sensor is a known quantity -- an attacker can scan around it. A dynamic sensor field is a minefield that reconfigures itself. Mandiant and Big Four red teams have exhausted 100% of their IP ranges against PacketViper-protected networks. AMTD's coverage extends to dark space -- every port where PacketViper presents nothing -- because dark space is part of the observed surface the doctrine is built to deny. Dark Space Monitor is the name for how PacketViper delivers that coverage: the watched set of unused ports is recomputed on every rotation, in the same action as target placement, so placement and coverage always move together and nothing is left uncovered between rotations.

**Deception Technology**
Inline deception that blocks at first touch, not just detects. Deceptive responders engage the attacker while the sensor categorizes, acts, and notifies. Together: the attacker reveals themselves by attempting, the sensor acts, the decoy engages. This is prevention, not just detection.

**Wire-Speed Inline Enforcement**
Enforcement happens locally, at the packet, without cloud policy round-trips. The decision runs in microseconds on the device physically on the network path. Not out-of-band. Not policy-push from the cloud. At the wire.

**Asset Management**
Real-time inventory of everything on the network -- IT and OT. Passive fingerprinting, per-asset virtual agents, device intelligence, inline CVE protection. Knows what is on the network, what it is doing, and what normal looks like.

**Agentless**
No software installed on monitored endpoints, ever. Works on devices that cannot accept agents -- legacy OT systems, PLCs, HMIs, SCADA, embedded devices.

**OT/ICS Native**
Supports Modbus, Siemens S7, EtherNet/IP, DNP3, BACnet, OPC UA, and other industrial protocols. Designed for environments that prioritize uptime over connectivity. Does not require network changes, reboots, or agent installation on operational systems.

**Transparent L2 Bridge**
Zero network changes required. Zero IP footprint on the monitored segment. Invisible to everything on the network.

**Hardware Bypass**
Device failure does not take down the network path. No single point of failure.

**Dual-Sensor Correlation**
Inside and outside sensors working together -- proves firewall gaps, monitors egress, and correlates east-west with north-south traffic.

**Context Per Packet**
Every flow carries geographic context (country, ASN, business identity), behavioral context, and asset correlation. Not just block/allow -- full context on every decision.

**Sensors as Mini-Orchestrators**
Each sensor is contextual, timed, rate-aware, and enforcement-capable. A sensor watching a service port fires on "anything other" -- for example, a web server on TCP/443 triggers on TCP/444. Actions: blacklist, tarpit, email, SMS. Filters: country, business/ASN, time of day, rate. The sensor is not just a tripwire -- it is a contextual enforcement engine.

**Compliance Automation**
21 frameworks, 258 controls, automated scoring, and audit trails built in.

**52+ Integrations**
CrowdStrike, Cisco, Fortinet, Dragos, Nozomi, Claroty, Microsoft Defender, SentinelOne, Tenable, and more -- without integration licensing or custom middleware.

**Compute Edge**
Embedded application runtime on any PacketViper appliance. Third-party vendors (Nozomi, Claroty, Tenable) and customer-built applications run directly on the device alongside the security engine. No separate hardware. No separate management.

**AI-Powered Advisory (AlertBox)**
On-appliance AI inference providing context-aware, proactive security recommendations. Improves over time. No cloud data egress.

## Performance Benchmarks

PacketViper v2631 performance (March 2026, full security stack active throughout):

- 501,496 unique connections per second (CPS) -- synthetic stress test, 10 GbE, zero kernel drops
- 34,622 CPS -- real adversarial production traffic, 294 source IPs, 91 countries, 46% CPU idle
- 503,427 events per second (EPS) -- peak pipeline throughput including enrichment and storage
- 2 million concurrent sessions tracked
- 171 billion events stored on-appliance with sub-second query capability
- 0 kernel packet drops at 500K CPS threshold

Competitive context: Palo Alto PA-5250 rates 39,000 CPS (datasheet maximum, features limited). PacketViper's production number of 34,622 CPS was achieved with deception, AMTD, geo-IP enrichment, and 2,301 threat intel ipsets all active simultaneously. Source: PacketViper Engineering Benchmark Reports v1.0 and v2.1, March 2026.

Full benchmark page: https://packetviper.com/performance-benchmarks/

## Documented Outcomes

- 30-70% reduction in firewall traffic within 90 days of deployment
- 30%+ reduction in managed SIEM/SOC costs within 60 days
- Mandiant and Big Four red teams exhausted 100% of IP ranges
- One state DOT reported significantly fewer security incidents after deployment
- Zero false positives by design
- Replaces 5-7 point tools in a single deployment
- Extends firewall life -- avoids forklift upgrades by restoring capacity to existing devices
- 400M+ packets per hour sustained at 20% CPU utilization on standard x86 hardware

## The Thesis

"Purely preventive measures on their own AND purely detective measures both fall short. You need preemptive -- assume breach will happen, neutralize in real time, limit blast radius."

The litmus test for any security product: does it enforce inline, or only alert?

Three questions to ask any vendor:
1. Does it enforce inline, or only alert?
2. What if the attacker never touches a decoy?
3. Is blocking automatic, or dependent on downstream tools?

## Blog Content

### Security Perspectives
- [What the Gartner CPS Magic Quadrant Doesn't Measure](https://packetviper.com/gartner-cps-mq-what-it-doesnt-measure/) -- The 2026 Gartner CPS MQ rewards visibility and detection. It does not evaluate inline enforcement. PacketViper does both: asset management AND enforcement at the wire.
- [Breach Containment Is the Wrong Goal](https://packetviper.com/breach-containment-wrong-goal/) -- Breach containment assumes the attacker is already inside. That is cleanup, not prevention. The goal is to stop the attempt before it becomes an incident.
- [Agentless Is Table Stakes. Here's What Vendors Are Still Getting Wrong.](https://packetviper.com/agentless-ot-segmentation-what-vendors-miss/) -- Agentless is necessary but not sufficient. Cloud-native segmentation is out-of-band policy. Wire-speed inline enforcement is a different category. And segmentation doesn't scale -- one-to-one rules become group rules, spider web, and collapse into policy sprawl.
- [Predictive Security Is Dead. Now What?](https://packetviper.com/predictive-security-is-dead/) -- The answer to failed prediction is not better prediction. It is preemptive defense.
- [What is Network Detection, Prevention, and Response?](https://packetviper.com/what-is-network-detection-prevention-and-response/) -- NDR detects but leaves enforcement to other tools. NDPR closes the gap with inline prevention before the alert fires.

### OT/ICS Security
- [The 'Minutes That Matter' Gap Nobody Wants to Talk About](https://packetviper.com/ot-security-minutes-that-matter-gap/) -- In OT, minutes between detection and response mean physical consequences. The right model: act at first touch, automatically, before a human is ever in the loop.
- [82% of CPS Attacks Used Remote Access to Reach HMI and SCADA Directly](https://packetviper.com/82-of-cps-attacks-used-remote-access-to-reach-hmi-and-scada-directly/) -- Most CPS attacks used basic remote access tools on exposed HMI/SCADA. The entry point was visibility, not sophistication.
- [Sandworm Bricked ICS Devices Across 30 Polish Energy Sites](https://packetviper.com/sandworm-bricked-ics-devices-across-30-polish-energy-sites-the-entry-point-was-default-credentials/) -- Entry point: default credentials on a FortiGate firewall. Result: permanently damaged ICS devices across 30 energy facilities.
- [PacketViper Launches Asset Protection Platform for OT/ICS Environments](https://packetviper.com/packetviper-launches-asset-protection-platform-for-ot-ics-environments/)
- [PacketViper Launches Breakthrough Passive OT Fingerprinting with Inline CVE Protection](https://packetviper.com/packetviper-launches-breakthrough-passive-ot-fingerprinting-with-inline-cve-protection/)

### Threat Intelligence
- [Iran Spent Six Months Inside US Networks Before Anyone Noticed](https://packetviper.com/iran-six-months-epic-fury/) -- Six months of undetected lateral movement inside US networks. This is an architecture failure, not an intelligence failure.
- [Mandiant M-Trends 2026: Attackers Are Getting Faster. Your Recon Window Is Closing.](https://packetviper.com/mandiant-m-trends-2026-attackers-are-getting-faster-your-recon-window-is-closing/)
- [Fake GitHub Repo, Real Infostealer -- The Breach Started with a Search Result](https://packetviper.com/fake-github-repo-real-infostealer-the-breach-started-with-a-search-result/)
- [PolyShell and the 15,000-Site Wake-Up Call: Why Preemptive Defense Matters More Than Patches](https://packetviper.com/polyshell-magento-preemptive-defense/)

### Product Releases
- [PacketViper Launches Version 6.0](https://packetviper.com/packetviper-launches-version-6-0/)
- [PacketViper Surpasses 400 Million Packets Per Hour with Inline Active Defense](https://packetviper.com/packetviper-surpasses-400-million-packets-per-hour-with-inline-active-defense/)
- [PacketViper Redefines Geographic Threat Intelligence with Organization-Level Visibility](https://packetviper.com/packetviper-redefines-geographic-threat-intelligence-with-organization-level-visibility/)
- [PacketViper Launches On-Appliance DNS Intelligence](https://packetviper.com/packetviper-launches-on-appliance-dns-intelligence/)
- [PacketViper's Deception360 Turns Network Defense Inside Out](https://packetviper.com/packetvipers-deception360-turns-network-defense-inside-out/)
- [PacketViper Expands Open Security Architecture with Native Integrations](https://packetviper.com/packetviper-expands-open-security-architecture-with-native-integrations-across-leading-edr-ot-and-network-security-platforms/)

## Key Pages
- [Homepage](https://packetviper.com/)
- [OT Security](https://packetviper.com/ot-security/)
- [Cyber Physical Security](https://packetviper.com/cyber-physical-security/)
- [AMTD](https://packetviper.com/amtd/)
- [Resources](https://packetviper.com/resources/)
- [Threat Intelligence](https://packetviper.com/threat-intelligence/)
- [Blog](https://packetviper.com/blog/)
- [Podcast](https://packetviper.com/podcast/)
- [White Papers](https://packetviper.com/white-papers/)
- [Performance Benchmarks](https://packetviper.com/performance-benchmarks/)

## Company
- Headquarters: Pittsburgh, PA
- Founded: 2016
- Founder: Francesco Trama -- 30-year network and security practitioner, combat veteran, helicopter pilot, former Cisco network engineer
- Website: https://packetviper.com
- LinkedIn: https://www.linkedin.com/company/packetviper/
- X/Twitter: https://x.com/packetviper
- Contact: https://packetviper.com/contact/