返回提示词目录第 1320 页

威胁指标筛选与封禁清单生成

筛查威胁指标的可信度与误封风险,生成历史日志检索方案、可导入名单及回滚记录。

完整提示词

<context>
Pushing every indicator from a report straight into every control is how teams block a cloud provider's shared IP, a CDN, a public file-sharing domain or a Microsoft binary hash, and then spend a day on rollback. Indicators also age at very different rates: attacker IPs change in days, domains in weeks, file hashes change with one rebuild, while behaviour lasts longer. Each type belongs in the control that can act on it: domains in DNS filtering and the proxy, full URLs in the proxy, IPs at the egress firewall only when they are dedicated infrastructure, SHA256 hashes in EDR, senders and attachment hashes in the mail gateway. Blocking before searching history also loses information: the more important question is who already talked to that infrastructure. Every block needs a source, an owner, an expiry and a way back.
</context>

<task>
Turn these indicators into blocks:

<indicators>
{{indicators}}
</indicators>

<controls>
{{controls}}
</controls>
{{#source_context}}

<source_context>
{{source_context}}
</source_context>
{{/source_context}}

1. Indicator review: normalise each indicator (refang only inside the final block lists, lowercase domains, SHA256 preferred, CIDR notation for ranges), identify its type and note its source and age.
2. Drop or downgrade indicators that would cause collateral damage or are weak: shared hosting, cloud and CDN IPs, popular legitimate domains or file-sharing services (block the specific URL instead, if the proxy allows), private address ranges, hashes of common legitimate files, and very old IPs. Say why for each.
3. Retro-hunt first: list the searches to run in DNS, proxy, firewall, mail and EDR logs for the past 30 to 90 days for every kept indicator, and say that hits become an investigation before or alongside the block.
4. Block plan: map each kept indicator to a control and an action (block, alert only, or monitor), with the reason. Alert-only is right where confidence is moderate or collateral is uncertain.
5. Produce the lists in a plain, importable shape per control (one value per line or a CSV with indicator, type, action, source, expiry, ticket), noting any vendor-specific format to check.
6. Change record: what changes, who approved, how to test (resolve a test domain, check a proxy log entry), and how to roll back.
7. Expiry and review: a default lifetime per type, for example 30 days for IPs and 90 days for domains, as starting points, plus who reviews.
</task>

<constraints>
- Never block an IP or domain without saying whether it is shared or legitimate infrastructure; if you do not know, flag it for checking rather than guess.
- Do not invent context about the indicators (malware family, actor) that is not in the input.
- Keep indicators defanged everywhere except the import lists, and label those lists clearly.
- Changes to production controls need the owner's approval; present the plan, do not imply it was applied.
{{> output/uncertainty}}
</constraints>

<output_format>
## Indicator review
Table: Indicator (defanged) | Type | Source | Age | Keep?

## Dropped or downgraded
Bullets with reasons.

## Retro-hunt first
Table: Log source | Search | Window.

## Block plan
Table: Indicator | Control | Action | Reason, followed by one fenced block per control with the import list.

## Change record
Bullets: summary, approver, test, rollback.

## Expiry and review
Table: Type | Default expiry | Reviewer.
</output_format>

可替换内容

  • {{indicators}}
  • {{controls}}
  • {{#source_context}}
  • {{source_context}}
  • {{/source_context}}
  • {{> output/uncertainty}}

分类与标签

编程开发、威胁情报、封禁清单、日志回溯、误封控制、变更回滚

前往 AI 图片创作