Stamp Banner Hashes and RDAP Diffs Into Shift Notes Before Weekend Containment Hands Off

By IPThreat Team November 17, 2026

Friday evening, a Tier-2 analyst opened a high-severity ticket for an ESXi management IP talking outbound on an uncommon TLS port. Reputation feeds labeled the address medium risk. Passive DNS showed two hostnames that looked like throwaway VPS labels. The analyst blocked the /32, pasted three vendor scores into the ticket, and signed off.

Saturday morning, GhostFrame-style ESXi implant chatter and a fresh wave of set-top proxy exits hit the same region. The weekend crew widened the block to the parent /24, then discovered half the prefix belonged to a shared cloud neighbor running customer VPN concentrators. Helpdesk tickets spiked. The original callback host had already renumbered. Nobody could reconstruct which OSINT facts were true at first contact versus which were guessed after the prefix widen.

The failure was incomplete IP witness work under shift pressure. Reputation alone travels poorly across handoffs. What travels is a short, timestamped record of what the address looked like when you first touched it.

What belongs in the first fifteen minutes

Treat every alert IP as evidence that will age out. Capture facts that expire or mutate before you open the containment debate.

  • Exact observation time in UTC, source sensor, and direction (ingress probe vs egress callback).
  • Service banners and TLS material: JA3/JA4 if you have them, certificate serial and SANs, HTTP Server header, SSH software string, and a hash of the raw banner bytes.
  • RDAP snapshot: registrant org, abuse contact, network name, allocation event date, and last change timestamp if the registry exposes it.
  • Route origin: announcing ASN, as-set membership if visible, and whether the prefix looks anycast or densely shared.
  • Neighbor context: two or three adjacent addresses on the same /24 that answer the same ports, plus whether those neighbors also appear in your honeypot or auth logs.

Write those five blocks into the ticket before you change firewall policy. Later analysts can still argue severity. They should not have to reconstruct what the host presented at 17:40 UTC.

Why this matters in the current noise mix

Recent reporting keeps stacking unrelated IP classes into the same queues. GhostFrame and Settra-linked ESXi implants push defenders toward admin-API and callback hunting. Parallel coverage of a botnet that turned roughly 1.3 million TV boxes into proxies floods geo and reputation systems with residential-looking exits. Separate threads on European refrigeration and building-control abuse add OT probes that share hosting ASNs with ordinary scanners.

Those campaigns do not share operators. They share cheap address space and overloaded analyst workflows. An IP that looks like ransomware staging infrastructure on Friday can resolve as a consumer proxy exit on Saturday if you only kept the vendor score. Banner hashes and RDAP diffs keep the classes apart when headlines collide in one SIEM view.

A practical OSINT order that survives handoff

1. Freeze the live face of the host

From a controlled jumphost, collect only what your policy allows: TCP connect plus banner grab on the alerted port, TLS cert dump if applicable, and a second check on 80/443 when the alert was non-web. Hash the outputs. Store them as ticket attachments named with the IP and UTC stamp.

Tradeoff: active probing can tip operators or trip upstream abuse desks. Keep probes minimal, rate-limited, and logged as defender traffic. If legal or peering constraints forbid outbound recon, fall back to your own sensor captures and note the gap explicitly.

2. Diff RDAP against yesterday, not against intuition

Pull current RDAP and compare it to any prior enrichment your TIP already stored for that prefix. Flag org renames, abuse-mailbox swaps, and allocation dates younger than 30 days. Fresh reallocations after a quiet period often mark bulletproof or reseller churn more reliably than a single bad reputation score.

Caveat: privacy redaction and RIR thin vs thick models hide contacts. Missing registrant data is not proof of malice. Record the redaction itself so the next shift does not re-query expecting a name.

3. Bind the route, then decide the blast radius

Confirm the origin ASN and whether the prefix is densely shared. Dedicated attacker VPS blocks often show long-lived origin with short-lived customer RDAP rows. Residential and set-top proxy exits often sit in carrier or CDN-adjacent space where /32 blocks punish bystanders and /24 blocks punish partners.

For ESXi callback suspects, prefer host-scoped controls plus process and API hunts on the hypervisor estate. For proxy-exit noise tied to TV-box style residential networks, prefer auth-step friction and session invalidation over wide network denies. Same ticket queue, different radius.

4. Cross-check neighbors without merging campaigns

If three neighbors share identical SSH banners and cert SANs, document a likely common image or scanner fleet. If only the alert IP matches your GhostFrame-related port and TLS profile while neighbors look like ordinary broadband CPE, keep the containment narrow and say so in the notes.

This step stops weekend crews from treating every coincident prefix hit as one actor after reading a ransomware headline and a proxy-botnet headline in the same shift brief.

Worked mini-scenario

Alert: 203.0.113.44 egress from an ESXi host to TCP/443 with an odd ALPN set. First-shift OSINT shows:

  • Banner hash H1 for a custom TLS stack; SANs list a single numeric hostname.
  • RDAP org changed 11 days ago; abuse mailbox points to a ticket-only reseller.
  • Origin AS is a known VPS provider; /24 also hosts unrelated customer VPN gear.
  • Two neighbors answer OpenSSH 8.x with stock Ubuntu banners; neither appears in hypervisor logs.

Decision recorded in the ticket: block 203.0.113.44/32 at the datacenter edge, isolate the ESXi host, hunt GhostFrame-like persistence and ransom staging tools on peer hypervisors, do not cap the full /24. Attach H1, RDAP JSON, and BGP origin. Weekend crew inherits a bounded plan instead of a reputation screenshot.

Contrast that with a helpdesk auth failure sourced from an address in a consumer ISP pool after the TV-box proxy takedown wave. Same OSINT checklist shows carrier RDAP, no custom TLS stack, high neighbor diversity, and reuse across unrelated customer logins. Response shifts to MFA step-up and credential reset sequencing. The IP still goes in the witness sheet so hunts can find infected internal clients that called the same proxy earlier in the week.

Implementation details that keep the habit alive

  1. Add five required fields to the IP enrichment form in your case system: observation UTC, banner hash, cert fingerprint, RDAP last-change, origin ASN.
  2. Auto-attach TIP snapshots on ticket create, then require a human “diff note” before severity can move above Medium.
  3. Store probe scripts and allowed ports in a defended runbook; forbid ad-hoc mass scans from analyst laptops.
  4. On shift change, read the witness block aloud or in chat before discussing wider ASN actions. If the block is missing, rebuild it before widening scope.
  5. Retire /32 blocks on a timer unless hunts confirm ongoing callback. Stale host blocks create a false sense that the campaign ended when the operator simply moved one address over.

Tradeoffs and caveats

Witness-heavy OSINT slows the first block by a few minutes. That cost buys fewer wrongful /24 quarantines when shared cloud and carrier space sits next to attacker VPS rows. It also fails open if analysts paste vendor scores and skip hashes under fatigue, so make the fields mandatory in software rather than optional in culture.

Open-source and commercial OSINT platforms disagree on first-seen and confidence. Your banner hash and RDAP diff are local ground truth for your ticket. External scores remain useful as secondary signals, not as the handoff artifact.

Legal and retention rules may limit how long you keep full banners or WHOIS personal data. Keep hashes and org-level RDAP where personal contacts must be purged, and note the retention class in the ticket.

Takeaways for the next weekend page

  • Start every IP investigation by freezing banners, certs, RDAP, and origin ASN with UTC stamps.
  • Separate ESXi/ransomware callback hosts from residential proxy exits before choosing blast radius.
  • Use neighbor similarity to support common imaging, not automatic campaign merge.
  • Hand off witness sheets, not reputation screenshots, when headlines about implants, proxy botnets, and OT probes land in the same queue.
  • Prefer timed host blocks plus estate hunts over prefix panic when shared tenants share the ASN.

IP OSINT earns its keep when the next analyst can replay what the address looked like at first contact. Stamp the mutable facts early, then contain with a radius that matches the host class you actually observed.

▲ Contact IPThreat