CodiLime
← Back to Capabilities

Capability

Security

Security is present at every layer, by design, from the first step to AI-ready. CodiLime builds protection into the data plane, automates the firewall estate, finds threats in network data, and keeps the AI agents you deploy inside their authorized scope. Our teams work across hardware, network, platform and software, so security keeps pace with every step of a modernization.

Measured impact

What we deliver

  • Data-plane security at line rate

    Inline inspection, AI web application firewalls and flood protection on DPUs and GPUs, so threats are stopped before they reach the host.

    Read the case study →

  • Threat detection in network data

    Machine learning over flow data that points at the addresses worth blocking, with results checked against threat intelligence.

    Read the case study →

  • Firewall automation across the estate

    Upgrades, migrations and policy changes for NGFW fleets, automated and run in maintenance windows at the scale of thousands of devices.

    Read the case study →

  • Security for the AI agents in your network

    Identity, scoped tool access, device-level permissions and a full audit trail for MCP-connected agents acting on live infrastructure.

    Read the case study →

  • Security products, built end to end

    Engineering for security vendors: an XDR platform from UX to cloud, east-west visibility and automated policy for a service mesh, and privileged access management.

    Read the case study →

How we do it

  1. Hardware & data plane

    Where traffic is inspected first

    DPU and GPU offload · Inline inspection · AI WAF · DDoS / SYN-flood · ACLs

  2. Network & perimeter

    The estate that enforces policy

    NGFW fleets · Upgrades and migrations · VPN and IPsec · Segmentation

  3. Platforms & workloads

    What runs between the services

    Kubernetes · Service mesh · East-west policy · Privileged access

  4. Detection

    Signals turned into decisions

    Flow analytics · ML classification · Threat intelligence · Anomaly detection

  5. AI agents

    The newest actors on the network

    Identity · Role-based tool access · Device-level permissions · Audit

  6. Across every layer

    Proven in a lab first

    Changes are exercised on emulated or physical labs before they touch production.

    Automated and auditable

    Every change goes through code, review and a record of who did what.

Technology

  • Data-plane acceleration

    • NVIDIA BlueField DPU
    • DOCA
    • NVIDIA GPU
    • CUDA
    • DPDK
    • P4

  • Firewalls & network security

    • Palo Alto Networks
    • Fortinet
    • Check Point
    • IPsec
    • OpenVPN

  • Identity & access

    • Keycloak
    • Open Policy Agent
    • OpenBao
    • TACACS+

  • Platforms & service mesh

    • Kubernetes
    • Istio
    • Envoy Proxy

  • Automation & AI

    • Ansible
    • Terraform
    • Python
    • Go
    • PyTorch
    • MCP

Proof of capability

Why CodiLime

  • Security engineers who build the software too

    The same teams cover the network, the platform and the code, so a security control is designed, built and automated in one conversation. That is how a firewall upgrade becomes a pipeline and a WAF becomes a DPU application.

  • Proven where a mistake is expensive

    Healthcare firewall fleets, a telecom operator’s traffic, an AI data center’s dataplane, and the products of security vendors themselves. Our security work runs in environments where downtime and breaches carry a real price.

  • Compliance built into the delivery

    Pipelines, access and environments are built so they can be audited. For Selector we built the delivery platform and supported its path to SOC 2 and ISO 27001 certification.

Insights

FAQ

Good to know

  • How do you inspect traffic at 100 Gbps without taking CPU from the hosts?

    By moving the inspection onto the network card. For an AI data center, CodiLime built an inline security dataplane on NVIDIA BlueField DPUs that reaches 100 Gbps per port with 0% host CPU tax and no agent on the host. The same approach carries an AI web application firewall that runs entirely on a GPU and a DPU, inspecting HTTP traffic at line rate and blocking threats before they reach the host kernel.

  • Can machine learning find malicious traffic in flow data?

    Yes, and the result can be measured. For a telecom operator, CodiLime built an ML pipeline over more than 2 TB of netflow. Of the IP addresses it flagged, 60% were confirmed malicious in VirusTotal within one week, which gives a security team a short, ranked list to act on.

  • How do you keep AI agents with network access inside their authorized scope?

    Each agent gets an identity, and every tool call and every device it touches is checked against policy. CodiLime built Net-Inspector to show this on a live network: role-based tool filtering, attribute-based device access, a lockout during maintenance windows, and an end-to-end audit of what each agent did, using Keycloak, Open Policy Agent and OpenBao.

  • How quickly can a large firewall fleet be upgraded?

    With the upgrade automated, the limit becomes the maintenance windows. CodiLime upgraded a fleet of more than 500 healthcare firewalls in less than three months, with up to 100 in a single window. For a firewall vendor’s Professional Services team, the automation we built and run upgrades more than 10,000 NGFWs and migrates more than 1,000 a year.

  • How do you see and control traffic between microservices?

    Through the service mesh, where every call between workloads already passes. For Mesh7, CodiLime co-designed and delivered a platform that gave security teams a real-time view of east-west traffic and enforced policy automatically, on the infrastructure they already ran.

  • Can you help a security or infrastructure product reach SOC 2 or ISO 27001?

    Yes, by building the platform so the evidence an auditor asks for already exists. For Selector’s AIOps platform, CodiLime built the automated, secure delivery pipeline across GCP, AWS, self-hosted setups and private clouds, and supported the company’s path to SOC 2 and ISO 27001 certification.

Tell us which layer worries you most.

The data plane, the firewall estate, or the AI agents about to get access to your network. CodiLime starts from your threat model and the stack you already run.