Observe before blocking
A new site starts in detection mode: everything is inspected and recorded, nothing is blocked. A report shows what would have been refused before you switch to blocking.
Readable detectors and rules, an anomaly score, modes per site, per rule and per path, narrow exceptions, virtual patches and a progressive rollout. You observe first, block later, and can always go back.
Where to find it in the console : Protection menu: Rules, Exceptions and Policies; Analysis menu: False positives; the Protection tab of a site.
A new site starts in detection mode: everything is inspected and recorded, nothing is blocked. A report shows what would have been refused before you switch to blocking.
Each event says which rule reacted, where (parameter, header, body), with which score and which action. No black box: you can replay a request and see the decision.
An exception targets one rule, one parameter, one path and one address, with an expiry date. The rest of the protection does not change.
Simulation on past traffic, rollout to a share of the nodes or of the requests, comparable versions and rollback in one step.
Every request goes through the same steps. The result is a decision, recorded with its evidence.
The built-in detectors (injections, deserialization, request smuggling, uploads, protocol), the OWASP CRS and your own rules look at the request. The catalogue documents more than 300 TW- error codes.
Each detection adds points according to its severity: 5 for critical, 4 for error, 3 for warning, 2 for notice. The default inbound threshold is 5: one critical detection blocks, two warnings (3 + 3) block, a single warning is only logged.
Off, detection or blocking, depending on the site, the rule, the path and the method. Some rules act on their own without going through the score, for example a fake Googlebot or an unsigned JWT.
The mode is set for the whole site, but also for one rule, a family of rules, a path or a method. A rule can also change its action: log, challenge, slow down, decoy, block or redirect.
The OWASP Core Rule Set v4 is evaluated by ToutWAF's own SecLang parser and evaluator, with the plug-ins that match the technology detected on your site.
The complete CRS v4 was only tested on a reduced fixture.
A readable condition language, with a visual editor and a text editor that stay in sync.
The console validator and the engine accept slightly different subsets of the language: always verify a rule with a test.
When a rule reacts wrongly, you work around it as narrowly as possible instead of switching it off.
Example of rule overrides, in the policy of a site
rule_overrides:
- { rule: tw-sqli-orderby-probe, mode: detect }
- { rule: "tag:attack-xss", mode: block, path_prefix: /checkout }
- { rule: "family:4xxx", action: challenge, methods: [POST] }Settings go from general to specific: global, then per organisation, per site and per path. The more specific one wins.
The recommended path is: observe, review, simulate, switch, watch, and keep the rollback ready.
You change the policy in a draft; nothing is published.
The policy is replayed on past traffic: requests that would be newly blocked or allowed.
A share of the nodes, a share of the requests or a single rule; you watch, then extend.
Publication is automatic and fast; if a node refuses the policy, it goes back to the last accepted version.
A known vulnerability is blocked at the WAF while the application waits for its fix.
ToutWAF settles its false positives and moves from observation to blocking, with evidence you audit and undo.
We prefer to say it plainly.
A week of detection, the report suggests a few narrow exceptions, you switch to blocking and keep the rollback ready.
A virtual patch blocks the exploit while the application waits for its update.
An exception limited to this field, this path and thirty days settles the case without touching the rest.
The basics are free; the advanced options need a paid offer. An option under licence stays visible in the console, locked, never hidden.
Voltar ao valor predefinido de uma opção avançada é sempre possível. O que acontece no fim de uma licença paga está descrito nas perguntas frequentes de preços: 48 horas de tolerância, 48 horas só de leitura e depois a proteção WAF é desativada até à renovação.
Instale-o no seu próprio servidor, ou contacte-nos para falar de licenças.