The Myth That Scanner First-Seen Dates Already Prove When Attackers Controlled the Address

By IPThreat Team November 16, 2026

On a Tuesday morning during a PHP gateway exploit wave, a regional MSSP pulled a single inbound IP from Check Point-facing decoy logs. Censys showed HTTPS open since March. Shodan listed the same banner family. The ticket owner wrote “attacker-controlled since Q1,” widened the block from /32 to /24, and briefed legal that the prefix had been hostile for eight months. By Thursday, a SaaS customer on that netblock lost partner SSO, the abuse mailbox filled with legitimate tenant complaints, and the real callback host for the implant still answered on a two-week-old VPS that never appeared in the first enrichment pane.

That failure sits next to the same week’s abuse traffic pattern: PHP2Shell-style probes against edge gateways, Macfinger ClickFix download hosts pushing macOS stealers, 0ktapus-style MFA phishing redirects, and Luna Moth extortion contact paths that reuse short-lived cloud IPs. Scanner first-seen timestamps feel precise. They measure when a crawler noticed a service, not when a threat actor leased, shared, or pivoted onto the address.

Where first-seen quietly misattributes

Internet-wide scanners record the earliest observation of an open port, certificate, or HTTP title from their vantage points. Hosting providers recycle addresses constantly. A clean marketing site can sit on 443 for months, get evicted, and leave the same IP to a bulletproof reseller within hours. Your enrichment still shows “first seen March,” which reads as long tenure under attacker control when the hostile use window started Monday night.

Residential and mobile CGNAT exits create a second trap. A ClickFix lure that phones home through a consumer ISP exit may share a /24 with thousands of unrelated subscribers. A first-seen date on port 8080 or a TLS cert on that exit describes network furniture, not campaign ownership. Extortion groups that email law firms a second time, as Luna Moth has, often rotate contact infrastructure faster than scanner indexes refresh. Treating the index clock as the actor clock inflates dwell-time estimates and steers containment toward the wrong neighbor tenants.

Signals that look like lease proof and are not

  • Open port longevity on 22, 80, 443, or 8443 after a generic cloud image boot.
  • Certificate notBefore dates that belong to a prior tenant’s Let’s Encrypt automation.
  • Passive DNS A records that still list marketing CNAMEs after the abusive hostname moved.
  • ASN org strings that name a hyperscaler while the hostile process runs inside one unpaid trial account.
  • VirusTotal “first submission” of a file hashed from that IP, which timestamps analyst upload behavior more than host control.

Rebuild the ownership clock with ticket-grade sources

Start from the alert IP, then force every enrichment field to answer one question: who could send packets as this address during the incident window? Keep scanner first-seen as a weak prior, never as the lease start.

Step 1: Pin live routing and registration to the hour

Pull current BGP next-hop and origin ASN from a looking glass or your own flow telemetry for the exact detection timestamp. Compare with RDAP/WHOIS netrange, organization, and last-changed fields. When IRR route objects disagree with live announcements, prefer live BGP for containment and park IRR mismatches as fraud or hijack follow-ups. During shared-cloud weeks, such as the TSMC-linked cloud provider breach chatter, assume multi-tenancy until you prove a dedicated allocation.

Step 2: Separate scout ingress from callback egress

PHP mass scouts that knock decoy ports after critical flaw notices rarely share addressing with the later implant C2. 0ktapus phishing kits often place lure pages, redirectors, and credential harvest APIs on different providers. Build two IP sets in the ticket: sources that touched your edge, and destinations your users or servers contacted afterward. Enrich each set independently. Merging them under one “campaign ASN” narrative collapses unrelated hosting customers into one quarantine.

Step 3: Time-box hosting tenure with at least three clocks

  1. Registration clock: RDAP update and netname change history for the /24 or more-specific.
  2. Service clock: Certificate Transparency issuance and HTTP title changes across consecutive days, not a single first-seen badge.
  3. Your clock: First and last hits in firewall, proxy, mail, IdP, and EDR for that IP during the incident.

When the three clocks disagree, write the disagreement into the escalation packet. Cursor’s confirmation that a supply-chain breach delivered a malicious Claude Code extension is a reminder that developer and SOC tooling can ingest hostile packages while IP reputation panels still look quiet. Ownership work has to survive vendor lag.

Step 4: Prove tenancy before widening

For cloud and CDN edges, ask whether the alert IP is anycast, shared LB, or a customer-assigned elastic address. Check reverse DNS patterns, HTTP Host headers from your logs, TLS SNI, and provider abuse portals that accept message-ID or URL evidence. One authenticated tenancy signal beats ten banner matches. Hold /24 and ASN escalations until you can name either a dedicated customer allocation or repeated hostile behavior across multiple tenants on that origin.

Real shifts that stress this workflow

Gateway exploit weeks. After PHP2Shell-class activity against Check Point gateways, scout IPs arrive in bursts from bulletproof and compromised WordPress hosts. First-seen dates on those scouts often predate the advisory by months because the hosts ran unrelated panels. Quarantine on observed exploit signatures and same-hour callback destinations, then backfill ASN velocity only when multiple distinct tenants show the same post-exploitation pattern.

MacOS ClickFix waves. Macfinger-style DownTheDRAIN delivery uses social engineering that lands users on short-lived download IPs. Those IPs may flip content between a parked page and a stealer drop within a scanner refresh interval. Snapshot the URL, hash, and responding IP at click time from proxy logs. Re-query OSINT an hour later as a change detector, not as proof the actor “owned” the address since the crawler’s first_seen.

Identity-centric intrusion sets. 0ktapus activity against large firm counts rides branded login clones and SMS or email lures. The interesting OSINT objects are redirect chains, kit fingerprints, and the small set of IPs that receive POSTed credentials. Envelope domains and CDN front doors distract. Bind IP investigation to the harvest endpoint and the residential or VPS exits that authenticate against your IdP within minutes of the lure click.

Extortion second passes. When groups such as Luna Moth return to law firms, contact addresses and payment sites churn. Prior case notes that pinned “the” operator IP from an old Shodan hit go stale. Rebuild from current message headers, sinkhole or sandbox callbacks, and payment-site hosting tenure measured in days.

Tradeoffs you should brief before changing playbooks

Deep RDAP and CT correlation adds ten to twenty minutes per priority IP. That cost pays off on admin portals, VPN concentrators, and partner SSO paths. It slows commodity scanner auto-blocks at the edge. Keep a two-track policy: automated /32 holds for clear exploit payloads, and human OSINT clocks before prefix or ASN actions that can strand customers.

Provider abuse tickets that request customer identity hit privacy and contractual walls. You often receive only “action taken” without a tenant name. Treat that as confirmation of shared risk, not as attribution. Peer ISACs and reciprocal sharing, including same-day notes around fragmented advisories, fill gaps scanners never will.

Historical datasets from commercial OSINT platforms remain useful for hunting recycled bulletproof ranges. Weight them as recycle risk, not as actor timeline. ISC Stormcast-style daily briefings help you mark which exploit families are hot this week so you know when scout noise will dominate enrichment queues.

Implementation details for the SOC desk

  • Add mandatory fields on IP tickets: live_asn, rdap_last_changed, first_internal_seen, last_internal_seen, scanner_first_seen, tenancy_status (dedicated, shared, anycast, unknown).
  • Reject escalation to /24 when tenancy_status is shared or unknown unless three unrelated internal victims contacted distinct hosts in that prefix within the same shift.
  • Diff CT and HTTP titles at T0 and T0+60m for every malware download IP tied to ClickFix-like lures.
  • Join IdP impossible-travel and MFA fatigue events to egress IPs only after carrier NAT and anycast checks clear.
  • Keep a one-page cheat sheet that states scanner first-seen equals crawler observation time, equals neither lease start nor exclusive control.

What to take into the next incident

Scanner first-seen dates belong in the weak-prior column. Live routing, RDAP change stamps, certificate churn, and your own log window build the ownership clock that survives PHP gateway scout floods, macOS stealer delivery hops, MFA phishing kits, and extortion rematches against professional services firms. Write the disagreement between those clocks into every escalation packet so containment width tracks tenancy evidence, not the comfort of a precise-looking timestamp from a global crawler.

▲ Contact IPThreat