A mid-shift SOC ticket lands with a single source IP hammering your VPN portal. Reputation feeds mark it high risk. GeoIP shows a residential ISP in a country where you have employees. Reverse DNS looks ordinary. The analyst blocks the /24, pages the ISP abuse desk, and closes the queue item. Ninety minutes later, the same credential-stuffing pattern resumes from seventeen other addresses on different consumer ASNs. The original block burned analyst time and a peer relationship on a rented exit node. The operator host stayed untouched.
That pattern is common in weeks when proxy markets and IoT abuse spike. After takedowns that turned more than a million TV set-top boxes into residential proxies, after listings that sell access to Chinese surveillance cameras, and during campaigns that park fraud domains at scale, many alert IPs are exits, not owners. OSINT for IP investigation has to classify the role of the address before containment widens.
What broke in the VPN portal case
The team treated a high reputation score as ownership. Consumer broadband prefixes rotate through botnet and proxy pools. A score that blends malware callbacks, spam traps, and open-proxy detections often fires on exits. Blocking at that layer stops one path for a few hours and leaves the staging host free to rotate.
They also skipped tenure and neighbor checks. The alert IP had appeared on three commercial blocklists in the prior forty-eight hours with the same /24 neighbors. Historic port banners from open scanners showed consumer CPE management ports open months earlier, then SOCKS and HTTP CONNECT listeners in the weeks before the VPN abuse. That timeline points to a compromised home device, not a leased bulletproof VPS.
A three-bucket model for every alert IP
Before you widen a block or open an abuse ticket, place the IP in one of three buckets. The bucket decides which OSINT steps matter and which containment action is safe.
- Exit: compromised residential, IoT, camera, or set-top device used as a proxy or VPN hop. Containment belongs on your edge (session kill, MFA step-up, rate cap). Abuse mail to the ISP is optional and low priority.
- Staging / operator host: dedicated or semi-dedicated hosting used for scanning, phishing kits, C2, or fraud-domain hosting. Prefix or ASN actions and hosting-desk contact become reasonable after you prove tenancy and role.
- Victim pivot: a customer, partner, or internal asset already compromised and used sideways. Containment is endpoint and identity work, not geo or ASN quarantine.
DoppelCart-style fraud networks that grow into hundreds of thousands of domains often share cheap hosting for the content plane and residential exits for the click and login plane. GhostFrame-style ESXi implants and Settra-linked ransomware traffic may show operator infrastructure on hosting ASNs while initial access probes arrive through proxy exits. GPUStack authentication-bypass exploit attempts in public telemetry likewise mix mass scanners on cloud prefixes with quieter follow-up from polluted residential pools. Mixing those planes in one block decision is how shifts lose hours.
Ordered OSINT checks that fit a live shift
Run these in order. Stop when the bucket is clear enough for an action. Perfect attribution can wait for the written case file.
1. Prefix tenure and list churn
Pull first-seen and last-seen dates from your threat intel platform or public scanners for the exact IP and its /24. Exits often show sudden multi-list appearance across many neighbors in a short window. Operator hosts more often show stable tenure on a hosting ASN with repeated service banners on the same ports.
Record three numbers in the ticket: days since first list hit, count of distinct lists in seven days, and how many /24 neighbors share those hits. High neighbor co-occurrence with short tenure leans exit. Low neighbor noise with long tenure on the same ASN leans staging.
2. Historic ports and banners
Compare open ports from at least two historical snapshots (for example ninety days ago and seven days ago). Consumer CPE and camera firmware leave fingerprints: TR-069, RTSP, vendor web UIs, default SSL cert subjects. A jump from those ports to SOCKS5, HTTP CONNECT, or atypical high ports supports proxy conversion. Hosting boxes used for fraud or exploit staging more often show stable HTTP/S, SSH, RDP, or panel ports with certificate subjects that match brand lookalike or panel hostnames.
When camera-access markets are active, treat open RTSP or vendor admin UIs on residential and small-business prefixes as strong exit signals even when reputation scores are still medium.
3. Reverse DNS, WHOIS org, and allocation type
Classify the allocation: consumer broadband, mobile CGNAT, business static, cloud, or hosting. CGNAT and consumer labels almost never justify ASN-wide quarantine from a single portal-abuse ticket. Cloud and hosting allocations invite deeper certificate and passive DNS work before you escalate.
WHOIS abuse contacts matter only after the bucket is staging. Sending urgent legal language to a residential ISP for a TV-box exit wastes the relationship you need when a real hosting prefix later hosts a phishing kit aimed at your brand.
4. Certificate and passive DNS for the hosting plane
If allocation type is hosting or cloud, pull current and historical certificates and passive DNS for the IP. Look for shared SANs across fraud-looking hostnames, short-lived certificates, and domain clusters that match kits rather than a single customer site. That evidence supports widening from one IP to a small prefix after you confirm the names are attacker-controlled content, not a shared CDN edge.
Shared CDN and large cloud prefixes need tenancy proof (HTTP host headers from your logs, SNI, or unique certificate) before any prefix cap. One probe IP on a mega-ASN is a weak campaign block seed.
5. Peer and path notes, not full attribution
A lightweight BGP or peer note helps when you already lean staging: same origin ASN as prior ransomware staging, sudden more-specific announcement, or path that matches bulletproof transit. Keep this as a supporting field. Do not let a peer graph alone upgrade an exit into an operator host.
Remediation that matches the bucket
Exit nodes
Kill the session, force step-up authentication, and apply short TTL rate limits on the targeted service. Hunt the username and device across other exits. Seed retrospective hunts with the blocklist hit time so you catch earlier quiet probes. Skip /16 and ASN blocks. Optional: low-priority ISP ticket with evidence that the CPE appears proxied, useful for metrics, not for same-hour relief.
Staging hosts
Cap the specific IP first, then the smallest prefix that shares certificate or passive DNS cluster evidence. Open the hosting abuse case with reverse DNS, cert SANs, first-seen scanner evidence, and your log excerpts. Hold ASN quarantine for repeated staging across multiple customers or for confirmed ransomware C2 after peer review.
Victim pivots
If the IP belongs to a known customer or partner range, or if reverse DNS and asset inventory match your environment, route to IR and EDR. Firewall blocks alone leave the compromised host free to rotate credentials.
Tradeoffs and caveats
Reputation-only triage is fast and wrong during proxy floods. Full OSINT on every scanner IP is slow and burns the queue. The three-bucket model with five ordered checks is the middle path: enough signal for a same-shift action without waiting for a finished attribution report.
GeoIP confidence drops when set-top and camera proxies spoil country labels. Use geo as a soft signal for auth distance alerts, not as the primary owner classifier. During Microsoft 365 or other vendor outages, separate availability noise from abuse traffic so OSINT time stays on IPs that actually authenticate or exploit.
Public scanner data lags live proxy rotation. Treat banners as directional. Confirm with your own edge logs (SNI, User-Agent, auth outcome, retry timing) before you widen containment.
Legal and abuse mail has a cost. Reserve urgent contacts for staging hosts with clear tenancy. For exits, your edge controls deliver the containment that matters to your users.
Shift checklist
- Tag the alert IP as exit, staging, or victim pivot before any prefix action.
- Capture tenure, neighbor co-list hits, and two historic port snapshots in the ticket.
- Confirm allocation type; keep residential and CGNAT on session and identity controls.
- For hosting or cloud, require certificate or passive DNS cluster evidence before widening beyond the single IP.
- Match containment to the bucket: edge controls for exits, scoped hosting caps plus abuse mail for staging, IR for victim pivots.
- Seed SIEM hunts with the first-seen list time so earlier callbacks surface while the shift is still open.
Proxy markets, camera-access sales, mass fraud-domain growth, and ransomware staging will keep mixing exit noise with operator hosts in the same alert queues. Teams that classify the IP role first keep containment sharp and save their escalation capital for hosts that actually run the campaign.