Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
eBPF runs small, verified programs inside the Linux kernel without loading a kernel module. For security teams it means detailed process, network, and file telemetry at a fraction of the usual cost and risk.

What Is eBPF?
Extended Berkeley Packet Filter is a mechanism for running small programs inside the Linux kernel, attached to events such as a system call, a network packet arriving, or a function being entered. The programs are checked by a verifier before they load, compiled to native instructions, and executed in kernel context. For security monitoring, that combination means you can observe what the kernel sees without the two traditional options: a kernel module that can panic the machine, or userspace polling that misses short-lived activity.
Classic Linux auditing through auditd works, and on busy systems it is expensive, difficult to filter precisely, and prone to dropping records under load. An eBPF collector filters and aggregates inside the kernel, so only the events you want cross into userspace. The performance difference on a high-throughput server is routinely an order of magnitude.
The safety argument matters just as much. An eBPF program that fails is rejected by the verifier or terminated; a faulty kernel module takes the host with it. That is not a theoretical distinction for anyone who has watched a security agent bring down a fleet, and it is the main reason vendors have been migrating their Linux sensors to eBPF since around 2021. The portability work known as CO-RE, compile once, run everywhere, removed the earlier requirement to compile against each kernel’s headers, which was the other adoption barrier.
execve./etc/shadow and writes to authorized_keys.setuid calls, capability changes, and namespace entry, which is how container escapes announce themselves.Falco, a CNCF project, is the best-known open detection engine built on this, shipping a modern eBPF driver and a rule language aimed at container and Kubernetes behaviour. Cilium Tetragon applies the same foundation with policy enforcement as well as observability, able to terminate a process in kernel context rather than after the fact. The bcc and bpftrace toolkits are invaluable for ad hoc investigation, and most commercial Linux EDR agents now use eBPF underneath whatever the datasheet calls it.
Compared with the Windows picture, eBPF plays roughly the part that Sysmon and the kernel callbacks play there: a source of structured, process-attributed telemetry that detections can be written against, which is why it belongs in any discussion of endpoint telemetry coverage.
The verifier imposes real limits: bounded loops, a ceiling on instruction count, and restricted memory access. Complex analysis belongs in userspace, with the kernel program doing filtering and aggregation. Older kernels are also a genuine constraint, since the useful security hooks arrived progressively, with BPF LSM landing in 5.7 and ring buffers in 5.8. Long-lived enterprise distributions backport selectively, so check the kernel rather than the distribution version.
There are also race conditions inherent to observing arguments at syscall entry, where a sufficiently careful attacker can modify memory between the kernel reading it and the program acting on it. Collectors mitigate this by reading at points where the data is already copied, but it is a reason not to treat eBPF telemetry as tamper-proof evidence.
The same capability serves attackers. Rootkits built on eBPF can hide processes, intercept data, and provide covert access, and several public examples exist. Loading a program requires CAP_BPF or root, so the defence is ordinary privilege control, plus monitoring the bpf() syscall itself. Alerting on unexpected BPF program loads is a short rule with an excellent signal-to-noise ratio, and it belongs on any Linux estate that has adopted the technology for defence.
At Privia Security, we assess Linux and container telemetry as part of our infrastructure and architecture review services, where the gap between Windows and Linux visibility is one of the most common findings.
For most purposes, yes. It performs far better under load, filters inside the kernel so less data crosses to userspace, attributes events to containers correctly, and avoids the record loss that auditd suffers on busy systems. auditd retains advantages where a mature, well-understood rule set already exists and where compliance requirements name it specifically, and some organisations run both during a transition.
A Microsoft-led open-source project brings eBPF to Windows, and it is maturing, but it is not the basis of Windows security telemetry today. On Windows the equivalent sources are kernel callbacks used by EDR products, Event Tracing for Windows, and Sysmon. Treat eBPF as the Linux and container answer, and plan Windows visibility separately rather than assuming one approach covers both.
Yes. Public rootkits built on eBPF can hide processes and files, intercept credentials, and provide covert remote access, and they are difficult to spot with conventional tooling. Loading a program needs root or the CAP_BPF capability, so the primary defence is privilege control. Add monitoring of the bpf syscall and alert on program loads that your own tooling did not perform.
Basic tracing works from around 4.9, but the security-relevant features arrived later: BPF LSM hooks in 5.7 and ring buffers in 5.8 are the ones most collectors want. Enterprise distributions backport selectively, so check the kernel and the enabled configuration options rather than the distribution release number, and confirm support on your oldest production hosts before standardising on a collector.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.