cboxdk/siem

A zero-runtime-dependency SIEM log-streaming core for PHP 8.4+. You produce one normalized security event; it hands you back exactly the bytes a real SIEM ingests — a Splunk HEC envelope, an Elastic ECS document, an ArcSight/syslog CEF line, a Graylog GELF message, or generic NDJSON. Nothing else: no HTTP, no queue, no credentials, no framework.

composer require cboxdk/siem

The two-tier picture

Log streaming to a SIEM splits cleanly into two jobs, and this package is deliberately only the first one:

1. Format — turn a normalized event into each SIEM's wire schema. Pure, deterministic, dependency-free, security-critical (CEF injection lives here). That is this package.
2. Deliver — ship those records over the network: SSRF-guarded egress, TLS, batching, a queue, retries, a dead-letter queue, encrypted secrets. That is the Laravel wrapper cboxdk/laravel-siem (separate package), plus a laravel-id audit binding. Not here.

Keeping the formatting core free of I/O means the security-sensitive part — the escaping that stops log injection — is small, framework-agnostic, and testable in isolation, and the delivery concerns live where a framework can do them properly.

Mental model

your app ──▶ SiemEvent ──▶ StreamFormatter ──▶ formatted record ──▶ StreamSink
           (normalize)      (per-SIEM schema)     (a string)         (deliver: wrapper)

• SiemEvent — one immutable, transport-neutral value object: a stable id, when it happened, an action, a category, an outcome, a severity, an optional actor and target, an optional source IP and message, and an already-flattened context bag.
• StreamFormatter — format(SiemEvent): string, one record per event. Ships five implementations. A batch is just the formatter mapped over many events; framing (NDJSON newlines, HEC concatenation, syslog envelopes) is the transport's job.
• StreamSink — a pure interface for delivery. The core ships no implementation (only Cbox\Siem\Testing\FakeStreamSink for tests); the real sink is the wrapper's.

Quickstart

use Cbox\Siem\Enums\EventCategory;
use Cbox\Siem\Enums\Outcome;
use Cbox\Siem\Enums\Severity;
use Cbox\Siem\Formatters\CefFormatter;
use Cbox\Siem\ValueObjects\Party;
use Cbox\Siem\ValueObjects\SiemEvent;

$event = new SiemEvent(
    id: 'evt_01HZX',
    occurredAt: new DateTimeImmutable(),
    action: 'user-login',
    category: EventCategory::Authentication,
    outcome: Outcome::Success,
    severity: Severity::Medium,
    actor: new Party('user', '42'),
    sourceIp: '203.0.113.7',
    message: 'User 42 signed in',
    context: ['method' => 'password', 'mfa' => true],
);

echo (new CefFormatter)->format($event);
// CEF:0|Cbox|SIEM|0.1.0|user-login|User 42 signed in|5|rt=... cat=authentication act=user-login ...

Swap the formatter for any other with no other change:

use Cbox\Siem\Formatters\EcsFormatter;
use Cbox\Siem\Formatters\SplunkHecFormatter;
use Cbox\Siem\Formatters\GelfFormatter;
use Cbox\Siem\Formatters\JsonFormatter;

(new EcsFormatter)->format($event);        // Elastic Common Schema JSON
(new SplunkHecFormatter)->format($event);  // Splunk HEC envelope
(new GelfFormatter('edge-1'))->format($event); // Graylog GELF 1.1
(new JsonFormatter)->format($event);       // generic single-line NDJSON

The formatters, mapped to each SIEM's real schema

Formatter	Target	Key schema facts
SplunkHecFormatter	Splunk HTTP Event Collector	{"time": <epoch-seconds.float>, "sourcetype": ..., "event": {...}}; time is seconds, not milliseconds; records concatenate as NDJSON.
EcsFormatter	Elastic Common Schema	@timestamp (RFC-3339 UTC), pinned ecs.version, event.{id,action,category[],type[],kind,outcome}, log.level, user.id, source.ip, labels.*, custom cbox.*.
CefFormatter	ArcSight / syslog	`CEF:0
GelfFormatter	Graylog (GELF 1.1)	version 1.1, host, short_message, epoch-seconds timestamp, numeric level 0–7, _-prefixed additional fields (_id forbidden).
JsonFormatter	generic / NDJSON	deterministic single-line, UTF-8 safe JSON.

Security posture (honest scope)

This package formats; it does not deliver. So its security surface is exactly one thing, and it takes it seriously: preventing log/record injection during formatting. The CEF formatter is the sharp edge — a CEF record is a single syslog line, so an unescaped |, =, or newline in attacker-influenced data could forge a header field, an extension key, or a whole second event. All escaping is isolated in a tested Cbox\Siem\Support\CefEscaper and proven with an adversarial round-trip test. Newline neutralization is unconditional — there is no config flag that can turn it off.

Everything downstream of a formatted string — SSRF-safe egress, TLS, auth, secret storage, retries — is out of scope here by design and belongs to cboxdk/laravel-siem. See SECURITY.md and docs/security/_index.md.

Requirements

• PHP 8.4+
• ext-json — ships with core PHP.

No runtime package dependencies. No framework. See docs/requirements.md.

Documentation

Full docs live in docs/: a quickstart, core concepts (the event model and the formatters), extension points (writing your own formatter), and the security posture and escaping guarantee.

Security reporting

Report vulnerabilities through GitHub Private Vulnerability Reporting — see SECURITY.md.

License

MIT — see LICENSE.