Enable use of Berkeley Packet Filter with seccomp
Requires the kernel to be built with CONFIG_SECCOMP_FILTER=y so processes can install BPF-based syscall filters (seccomp-bpf).
Checked against the resolved running state (e.g. sshd -T, sysctl, systemctl show), catches drop-ins and Includes a file read would miss. Caveat: runtime ≠ persistence; a value correct now may not survive a reboot.
Pavois asserts the effective configuration, the live, resolved state, not a file. File-based scanners (OVAL/SCAP, Lynis) miss Includes, drop-ins and runtime defaults; this check sees what is actually applied.
A mapping is a cross-reference to where each standard places this requirement, anchored and cross-validated, not a claim of equivalence. A passing check is evidence toward these references, how to read it.
Why this rule matters
Use of BPF filters allows for expressive filtering of system calls using a filter program language with a long history of being exposed to userland. CONFIG_SECCOMP_FILTER enables seccomp-bpf, the mode that lets a process install a fine-grained allowlist of syscalls; this is what container runtimes, systemd SystemCallFilter=, browsers and sandbox daemons rely on to shrink a compromised process's reach into the kernel, directly reducing the kernel attack surface exposed to untrusted code.
What Pavois checks
Pavois reads the build options of the running kernel, grep '^CONFIG_SECCOMP_FILTER=' /boot/config-$(uname -r) or zcat /proc/config.gz, and requires y. Reading the live kernel's own config (keyed on uname -r) reflects exactly the booted kernel, which a package-name guess or generic file cannot.
describe command("C=/boot/config-$(uname -r); if [ -r \"$C\" ]; then cat \"$C\"; elif zcat /proc/config.gz 2>/dev/null | head -1 | grep -q .; then zcat /proc/config.gz; else echo PAVOIS_NO_KERNEL_CONFIG; fi | grep -E '^(CONFIG_SECCOMP_FILTER=|PAVOIS_NO_KERNEL_CONFIG)'") do
its('stdout') { should_not match(/PAVOIS_NO_KERNEL_CONFIG/) }
its('stdout') { should match(/^CONFIG_SECCOMP_FILTER=y$/) }
endHow to verify it is applied
Run grep '^CONFIG_SECCOMP_FILTER=' /boot/config-$(uname -r) (or zcat /proc/config.gz | grep CONFIG_SECCOMP_FILTER). Expected output:
CONFIG_SECCOMP_FILTER=y
You can also confirm runtime support: grep Seccomp /proc/self/status exists and systemd-analyze reports seccomp filtering as available.
Inspect & investigate
No routine log entry. A syscall blocked by a seccomp filter (when set to log/kill) appears in the audit log (/var/log/audit/audit.log, type=SECCOMP) and in dmesg / journalctl -k. Confirm the build flag with grep CONFIG_SECCOMP_FILTER /boot/config-$(uname -r).
Remediation
This is a kernel build-time option, so Pavois ships no automated harden step for it. It is satisfied by booting a distribution kernel compiled with CONFIG_SECCOMP_FILTER=y (true for all current Debian/Ubuntu/RHEL kernels, required by Docker/Podman/systemd sandboxing); a custom kernel must be rebuilt.
Pavois applies this with its own harden engine, the plan below, not a shell script:
| resource | kernel_build |
|---|
pavois harden plan localwhere the target is local, a user@host SSH alias, or a container , Docs
Impact & precautions
If absent, processes cannot install syscall allowlists, so container and systemd sandboxing silently loses a key confinement layer and any compromised service has the full syscall surface. Precautions: the option only enables the capability; it does not block anything by itself, so turning it on (via a compliant kernel) cannot break running software. The risk is the opposite, leaving it off weakens isolation. No lockout concern beyond the normal kernel switch.
Standards mapping
| Standard | Reference | Type | Version | Confidence |
|---|---|---|---|---|
| ANSSI BP-028 | R20 | direct | 2.0 | high |
Each reference is a cross-reference anchored in the upstream benchmark and cross-validated against the SCAP Security Guide and ansible-lockdown, not a claim of equivalence. Direct = a prescriptive, line-level requirement; supporting = an abstract control family (NIST) the check provides evidence toward. How to read a mapping.