Web Security

Download 0
Last updated Nov 23, 2026

AI Agents

The Web Security app provides two AI agent templates. The Web Security Lead Analyst takes requests and hands URL template registration to its sub-agent, the Web Path Profiler.

Note
AI agents need the AI Assistant or Agentic SOC app and the matching license feature.

Installing the template

  1. Open the AI Agents menu of the AI Assistant (or Agentic SOC) app.
  2. Click Templates on the agent list.
  3. Install Web Security Lead Analyst. Its sub-agent, Web Path Profiler, is installed with it.

Only administrators can install templates. Updating the app does not change agents already installed. To bring one to the latest definition, delete it and install it again.

AI Assistant panel

With the agent installed, you can open the AI Assistant panel on the right of the Web Security screens. The panel talks to the Web Security Lead Analyst, so you can ask while you watch Web Monitoring.

Web Security Lead Analyst

It handles a request as one of two kinds.

  • Profiling: Registers or fixes URL templates and sets parameter types. It does not do this itself; it hands the work to the Web Path Profiler.
  • Anomaly analysis: Investigates the web logs already collected for bots, errors, object ID crawling, object enumeration (BOLA) and parameter anomalies itself. The conclusion gives each suspicious subject its evidence and recommendations.

When a request names no table or period, it lists the tables collected with the web log schema and asks you to pick.

Example requests

  • Register the URL templates of the web_access table and set the parameter types
  • Find IPs suspected of object enumeration on web_access in the last hour
  • Analyze what the IPs with the most errors did yesterday

Web Path Profiler

Investigates the web logs and registers URL templates and parameter types. It repeats these steps.

  1. Looks at request counts by host, method and path.
  2. Registers busy fixed paths and static resource directories first.
  3. Keeps only the requests no template normalizes, and groups them by path shape (numbers, UUIDs and so on replaced by tokens).
  4. For each top path shape, looks at real paths and registers a template with meaningfully named placeholders.
  5. Repeats 3 and 4 until the share of normalized requests (coverage) reaches the target.
  6. For each parameter, sets the type and the length range from the value count, subject count, values per subject, 4xx share and value lengths.
  • Scanning noise is not registered: paths answered mostly with 404, .php, /.env, /.git/ and the like.
  • Registering and removing are saved at once, without approval. When it removes a wrongly grouped template, it leaves the evidence in the result.
  • The result gives the numbers of templates added, updated and removed, the coverage, the parameters set to Object ID with the evidence, and the unknown paths left.

To run the Web Path Profiler directly, without the lead analyst, give it these inputs.

InputDescription
RequestWhat to do. For example: register the URL templates of web_access
Web log queryThe logs to investigate. For example log schema=web duration=7d source="logger GUID"
Target hostThe table name
Coverage target0.98 by default
Subject fieldsrc_ip by default. Use a user ID or session ID field when the logs have one

Query commands the agents use

CommandPurpose
websec-path-templatesList registered templates
websec-add-path-template-batchRegister or update the template of each input row
websec-remove-path-template-batchRemove the template of each input row
websec-normalize-path-batchNormalize request paths to templates
websec-parse-query-params-batchSplit a query string into keys and values
websec-extract-params-batchOutput one row per parameter of each normalized request
websec-classify-user-agent-batchClassify User-Agents into browser families and bot types
websec-score-bola-batchJudge BOLA against a baseline

See the app's Commands tab for the options and output fields of each command.