Posted in Cloud & Infrastructure

eBPF-Based Runtime Security Tools Explained: Falco vs Tetragon vs Aqua Tracee

Service meshes like Istio, Linkerd, and Cilium get a lot of attention for how they move and secure traffic between services. But traffic encryption and L7 policy don’t catch a process inside a container spawning a reverse shell, writing to a file it has no business touching, or escalating privileges once an attacker is already on the box. That’s a different problem — runtime security — and the tool of choice for solving it at scale is eBPF: a Linux kernel technology that lets lightweight, sandboxed programs run inside the kernel itself to observe or intervene in what’s actually happening on a host. Three open-source projects have become the standard names in eBPF-based runtime security: Falco, Tetragon, and Aqua Tracee. They overlap in purpose but differ in architecture and depth of enforcement.

At a Glance

ToolMaintainerCore ApproachEnforcementGovernance
FalcoOriginally Sysdig; CNCF graduated projectKernel syscall events + K8s audit logs + cloud events, checked against rulesDetection and alerting (response via Falcosidekick to 50+ destinations)CNCF graduated project (2024)
TetragonIsovalent (part of Cisco)eBPF kernel observability tied to the Cilium project’s toolingCan both observe and actively block/kill actions in-kernel in real timeOpen source, developed within the Cilium ecosystem
Aqua TraceeAqua Security (Team Nautilus)eBPF syscall + Linux Security Module (LSM) hook events, rule engine in Go/RegoDetection/forensics by default; enforcement available via Aqua’s commercial platformOpen source (aquasecurity/tracee on GitHub)

Short Verdict

Falco is the most broadly adopted and vendor-neutral choice, with the largest rule ecosystem and the deepest CNCF governance, making it a safe default for teams that mainly need detection and alerting across containers, hosts, and cloud events. Tetragon stands out when real-time, in-kernel enforcement matters — stopping a malicious process before it finishes executing, not just alerting after the fact — and it’s the natural choice for teams already running Cilium for networking. Aqua Tracee is the best fit for teams wanting a lightweight, portable, open-source forensics tool with LSM-hook-based detection that’s harder to evade than plain syscall tracing, especially if they’re already evaluating or using Aqua’s commercial platform.

What eBPF Is, and Why Runtime Security Tools Use It

eBPF (extended Berkeley Packet Filter) is a Linux kernel technology that allows small, verified programs to run safely inside the kernel without modifying kernel source code or loading traditional kernel modules. Before eBPF, observing low-level activity — every system call a process makes, every file it opens, every network connection it initiates — typically meant either slow, invasive instrumentation or a custom kernel module carrying real stability and security risk. eBPF programs are checked by an in-kernel verifier before they’re allowed to run, can be attached to specific kernel events (a system call, a network packet, a security hook), and execute with minimal overhead because the filtering and aggregation happen in the kernel instead of shuttling every raw event up to userspace first. That combination — safety, low overhead, and deep visibility — is exactly what runtime security needs: the ability to watch everything a workload does at the kernel level, in real time, without the workload itself being modified or aware it’s being observed. This is fundamentally different from a service mesh sidecar, which operates at the network layer between services; eBPF-based tools operate underneath the application entirely, at the level of the operating system itself.

Falco: The CNCF-Graduated Standard

Falco was created by Sysdig and donated to the Cloud Native Computing Foundation in 2018, reaching incubating status in 2020 and graduating — the CNCF’s highest maturity tier — in 2024. Falco describes itself as working like “a network of security cameras” for infrastructure: it collects events from the Linux kernel (via a modern eBPF probe, its preferred collection method, or an older kernel-module option), from Kubernetes audit logs, and from cloud event sources like AWS CloudTrail, then checks all of it against a large library of rules. Built-in and custom rules can flag behavior such as cryptomining processes, unexpected file exfiltration, privilege escalation attempts, and rootkit installation. Falco itself focuses on detection and alerting; its companion project Falcosidekick forwards matched events to more than 50 downstream destinations — Slack, email, message queues, serverless functions — so a response action can be triggered externally. Falco’s biggest advantages are its maturity, its large community-maintained rule set, and its CNCF governance, which makes it a low-risk default pick for most organizations.

Tetragon: Real-Time Enforcement From the Cilium Ecosystem

Tetragon comes out of Isovalent, the company behind the Cilium networking and security project; Isovalent was acquired by Cisco (announced in December 2023), and Isovalent’s own materials now identify the team as “Isovalent at Cisco,” operating as part of Cisco’s security portfolio. Isovalent describes Tetragon as providing “eBPF-based transparent security observability combined with real-time runtime enforcement.” The key architectural difference from Falco is that Tetragon’s policies run entirely in the kernel and can actively intervene — not just alert — stopping a dangerous system call or action as it happens rather than only logging it afterward. Because the filtering and policy logic execute in-kernel, overhead stays low even with fine-grained enforcement rules in place. Tetragon is open source and doesn’t require Cilium to run, but it shares engineering and community ties with the broader Cilium project, making it a natural complement for teams that have already adopted Cilium for Kubernetes networking and want the same eBPF foundation extended into runtime security enforcement.

Aqua Tracee: Forensics-Focused and LSM-Aware

Aqua Tracee, released by Aqua Security’s Team Nautilus in 2019, was originally built around eBPF syscall tracing but evolved specifically because syscall tracing alone is vulnerable to time-of-check-to-time-of-use (TOCTOU) evasion techniques, where an attacker changes what a syscall actually does between the moment it’s checked and the moment it executes. To close that gap, Tracee added collection of Linux Security Module (LSM) hook events, which Aqua describes as giving more comprehensive and reliable detection than syscall tracing by itself. Tracee’s architecture splits into an eBPF-based tracing component that collects events and a rule engine (supporting both Go and Rego/OPA policies) that evaluates them against built-in behavioral signatures — covering things like anti-debugging techniques, possible code injection, fileless execution, unauthorized kernel module loading, and LD_PRELOAD abuse. A 2021 update (version 0.6.0) added CO-RE (Compile Once – Run Everywhere) portability, letting Tracee run on BTF-enabled kernels without needing to compile the eBPF probe separately for each kernel version — a meaningful operational improvement for anyone running mixed kernel versions across a fleet. Tracee is fully open source on GitHub; Aqua’s commercial platform builds on the same eBPF foundation to add runtime enforcement and broader workload protection.

Trade-offs

  • Falco — Pros: largest community and rule library, CNCF-graduated governance, broad event-source support beyond just the kernel. Trade-offs: primarily detection/alerting rather than in-kernel blocking; response actions depend on integrating Falcosidekick with an external system.
  • Tetragon — Pros: genuine real-time in-kernel enforcement, not just alerting; low overhead by design. Trade-offs: smaller standalone community than Falco; closest value realized alongside the broader Cilium/Isovalent ecosystem, now under Cisco.
  • Aqua Tracee — Pros: LSM-hook detection resists simple syscall-level evasion; CO-RE portability simplifies multi-kernel deployments. Trade-offs: full enforcement capability sits behind Aqua’s commercial platform rather than the open-source core; smaller community footprint than Falco’s.

Who This Matters For

Security and platform teams running containerized workloads in production — particularly on Kubernetes, where a single compromised container can be a foothold into much more — are the core audience. These tools matter most for organizations that need to detect or stop in-host threats (privilege escalation, container breakout attempts, cryptomining, data exfiltration) that happen below the network layer a service mesh or firewall can see. They’re complementary to, not a replacement for, network-layer security tools like Cilium’s own network policies or Istio’s mutual TLS.

Bottom Line

All three tools rest on the same core technology — eBPF’s ability to observe kernel-level activity cheaply and safely — but they diverge on what happens next. Falco is the mature, CNCF-governed detection standard with the broadest adoption. Tetragon pushes further into real-time in-kernel enforcement and is backed by the team behind Cilium, now inside Cisco. Aqua Tracee carves out a niche with LSM-hook detection that’s more resistant to evasion than syscall tracing alone, with commercial enforcement available through Aqua’s broader platform. Many organizations end up running Falco or Tetragon as their primary detection layer and evaluating Tracee for specific forensics or evasion-resistance needs, rather than treating the three as mutually exclusive.

FAQ

What’s the difference between eBPF-based runtime security and a service mesh like Cilium or Istio?
A service mesh secures and observes traffic moving between services at the network layer. eBPF-based runtime security tools operate at the operating-system level, watching what processes inside containers actually do — system calls, file access, privilege changes — regardless of network activity.

Does using Cilium for networking require also using Tetragon for security?
No. Tetragon is a separate, independent project that doesn’t require Cilium to run, though it shares the same eBPF foundation and engineering lineage, and many teams adopt both together.

Is Isovalent still an independent company?
No. Cisco announced its acquisition of Isovalent in December 2023, and Isovalent’s own current materials identify the team as operating as part of Cisco.

Can these tools block an attack, or do they only alert?
It depends on the tool. Falco is primarily a detection and alerting system, relying on external integrations for response. Tetragon can enforce policy and block actions directly in the kernel in real time. Aqua Tracee’s open-source core is detection-focused, with active enforcement available through Aqua’s commercial product.

Do these tools require special kernel versions?
eBPF capabilities depend on the Linux kernel version and configuration (such as BTF support for CO-RE portability). Falco can also fall back to a kernel-module collection method on systems where its modern eBPF probe isn’t the best fit. Always check each project’s current documentation for supported kernel versions.

Sources

Unknown's avatar

Author:

Innovations, logic & design thinking