Clair vs Falco

Clair is free, self-hosted software that continuously scans container images for vulnerabilities in operating-system bases, installed packages, and programming-language components. Clients use its API to index images and match them against known vulnerabilities. Clair updates its vulnerability database as new issues appear and can reanalyze images without rereading their complete contents. It supports OCI Distribution and Docker v2 images, and lists RHEL, Alpine, Ubuntu, Debian, Oracle, Photon, and SUSE distributions, plus Java, Python, Golang, and Javascript. Deployment options include microservices, monolithic installations, CI pipelines, and disconnected environments. Clair can run as an indexer, matcher, notifier, or combo process. It uses PostgreSQL for persistence and is released as a Quay.io container, with the clairctl command-line tool among its artifacts. Clair runs on Linux, supports registry scanning, and is licensed under Apache 2.0. Runtime protection is not provided.

Falco is a free, open source runtime security project for Linux and self-hosted environments. It monitors hosts, containers, Kubernetes, and cloud environments for abnormal behavior, potential threats, and compliance violations in real time. Events are compared with built-in detection rules, and users can create rules for their own environments. Sources include Linux kernel system calls, Kubernetes audit logs, cloud events such as AWS CloudTrail, and events from GitHub and Okta. For kernel event monitoring, Falco uses an eBPF driver by default and offers a kernel module for wider compatibility. Installation options cover Linux hosts, Docker containers, and Kubernetes, including official Helm charts; supported architectures include x86_64 and ARM64. Falcosidekick can forward events to more than 50 systems, including email, chat, message queues, serverless functions, and databases. Falco is hosted by the CNCF as a graduated project. Detection depends on having appropriate rules in place, while resource use varies with server load and workload. The project recommends testing performance in the target environment.