Error (Hit/Block) Logs in JSON Format

[superstes]

Summary

I have not seen any option in the documentation to change the error/block log format.

JSON format would make sense for many use-cases. Also the audit-logs seem to already support it.

Basic example

It would be nice to get this as json object:
[client \"::ffff:95.214.55.x\"] Coraza: Warning. Host header is a numeric IP address [file \"/etc/coraza-spoa/coreruleset/rules/@owasp_crs/REQUEST-920-PROTOCOL-ENFORCEMENT.conf\"] [line \"1772\"] [id \"920350\"] [rev \"\"] [msg \"Host header is a numeric IP address\"] [data \"159.69.187.x\"] [severity \"warning\"] [ver \"OWASP_CRS/4.0.0-rc2\"] [maturity \"0\"] [accuracy \"0\"] [tag \"application-multi\"] [tag \"language-multi\"] [tag \"platform-multi\"] [tag \"attack-protocol\"] [tag \"paranoia-level/1\"] [tag \"OWASP_CRS\"] [tag \"capec/1000/210/272\"] [tag \"PCI/6.5.10\"] [hostname \"::ffff:159.69.187.x\"] [uri \"/\"] [unique_id \"FMPGEMUVBOHBCEMH\"]

Motivation

JSON is much easier to parse than the stringified format.
Log systems like Graylog and Grafana-Loki/-Agent can parse JSON natively & performant. That is very convenient - especially as such security-logs are very important to process.

Users could compile coraza with a error-callback, but that is only a workaround and not that easy/clean to implement when using a 'connector' like coraza-spoa

From what I've read into the source - this is where the logs are written: https://github.com/corazawaf/coraza/blob/main/internal/corazarules/rule_match.go#L254

Related to: #856, corazawaf/coraza-caddy#20, corazawaf/coraza-spoa#91, #1150

I'm open to contribute.

[superstes]

Of course - one could use WithErrorCallback to override it with a custom format - but that might not be doable or be favourable for 'users'.

[superstes]

Basic callback for reference:

func createWAF() coraza.WAF {
	waf, err := coraza.NewWAF(
		coraza.NewWAFConfig().
			WithErrorCallback(logErrorJSON),
	)
	if err != nil {
		log.Fatal(err)
	}
	return waf
}

type errorLogJSON struct {
	File       string   `json:"file"`
	Line       int      `json:"line"`
	ID         int      `json:"rule_id"`
	Revision   string   `json:"rev"`
	Msg        string   `json:"msg"`
	Data       string   `json:"data"`
	SeverityID int      `json:"sev_id"`
	Severity   string   `json:"sev"`
	Version    string   `json:"ver"`
	Maturity   int      `json:"mat"`
	Accuracy   int      `json:"acc"`
	Client     string   `json:"client"`
	Disruptive bool     `json:"disruptive"`
	Tags       []string `json:"tags"`
	Server     string   `json:"server"`
	URI        string   `json:"uri"`
	UniqueID   string   `json:"unique_id"`
}

func logErrorJSON(mr types.MatchedRule) {
	r := mr.Rule()
	j, _ := json.Marshal(errorLogJSON{
		File:       r.File(),
		Line:       r.Line(),
		ID:         r.ID(),
		Revision:   r.Revision(),
		Msg:        mr.Message(),
		Data:       mr.Data(),
		Severity:   r.Severity().String(),
		SeverityID: r.Severity().Int(),
		Version:    r.Version(),
		Maturity:   r.Maturity(),
		Accuracy:   r.Accuracy(),
		Client:     mr.ClientIPAddress(),
		Server:     mr.ServerIPAddress(),
		Disruptive: mr.Disruptive(),
		Tags:       r.Tags(),
		URI:        mr.URI(),
		UniqueID:   mr.TransactionID(),
	})
	fmt.Println(string(j[:]))
}

Results in:
{"file":"coreruleset/rules/REQUEST-941-APPLICATION-ATTACK-XSS.conf","line":1210,"rule_id":941160,"rev":"","msg":"NoScript XSS InjectionChecker: HTML Injection","data":"Matched Data: \u003cscript found within REQUEST_HEADERS:Referer: http://localhost:8090/?a=\u003cscript\u003ealert(1)\u003c/script\u003e","sev_id":2,"sev":"critical","ver":"OWASP_CRS/4.7.0-dev","mat":0,"acc":0,"client":"[::1]","disruptive":false,"tags":["application-multi","language-multi","platform-multi","attack-xss","xss-perf-disable","paranoia-level/1","OWASP_CRS","capec/1000/152/242"],"server":"","uri":"/favicon.ico","unique_id":"YuxjbYnedOyrPYNkyFy"}

[superstes]

BTW: Looks like the ocsf-auditlog does something similar - https://github.com/corazawaf/coraza/blob/main/internal/auditlog/formats_ocsf.go#L84

[fzipi]

This is normally handled by using the SecAuditLogFormat directive.