← Back to blog

Mapping spyware infrastructure from public records

A surveillance campaign leaves a paper trail in public logs long after its servers go dark. Starting from the Predator indicators published by MVT, the Amnesty Security Lab's toolkit project, here is how you read that trail without ever contacting the infrastructure.

Security11 min readBy Ryan SebastianUpdated

The rule before the method

Investigating surveillance infrastructure has one hard rule: you do not touch it. You do not resolve the domains against the live hosts, you do not connect to them, you do not scan them. The target may be live attack infrastructure, and a single connection can tip off an operator or, worse, look like interaction from the very people the campaign is hunting.

So the whole investigation runs on passive records: logs and registries that third parties keep, which you can read without the target ever knowing. The two that matter most are Certificate Transparency and domain registration data. This walk-through uses only those, and the result shows how much of a campaign you can reconstruct from the outside.

The starting point: published indicators

You rarely start from nothing. For mercenary spyware, the research community publishes indicators, and Amnesty International's Security Lab is one of the best sources. The indicator set its MVT project keeps for Predator, the spyware sold by the Intellexa alliance, lists 569 domains used across its operations, gathered from published research by several teams and from the Lab's own investigations. That is the seed.

Read the seed before you enrich it, because the names already tell a story. Grouped by what they imitate:

  • News outlets: almasryelyuom.com and almasrylayoum.com imitate Al-Masry Al-Youm, an Egyptian daily. efsyn.news imitates the Greek paper Efimerida ton Syntakton. dw-news.co imitates Deutsche Welle. africa-confidentiel.fr, afrinew.net, altsantiri.news and allafrika.live follow the same pattern.
  • Government and banks: eg-gov.org poses as an Egyptian government site, bank-alahly.com as the National Bank of Egypt.
  • Brands and lures: api-apple-buy.com for Apple phishing, covid19masks.shop for a pandemic-era lure.
  • Throwaway redirectors: short domains like 02s.co and 5m5.io, the kind used as the first hop in a one-click infection link.

Across all 569, the top-level domains skew cheap and disposable: 204 .com, 71 .xyz, 50 .net, 30 .live, 28 .co, and a long tail of .info, .me, .news, .online and .shop. This is the shape of a campaign that treats domains as consumables. That shape is itself the first finding.

Enrichment one: Certificate Transparency

Every time a web server gets a TLS certificate from a public authority, that certificate is logged, publicly and permanently, in Certificate Transparency logs. The system exists to catch mis-issued certificates, but for an investigator it does something else: it records when a piece of infrastructure was stood up, by which certificate authority, and under which names, and it keeps that record after the infrastructure is gone.

Querying the logs for a sample of the Predator domains (a passive query against the logs, not the domains) returns a clear pattern. Taking the Egyptian and Greek campaign domains:

  • almasryelyuom.com: 8 certificates, first seen June 2021, last expiring March 2022.
  • alraeesnews.net: 10 certificates, April 2021 to March 2022.
  • efsyn.news: 16 certificates, June 2021 to February 2022.
  • bank-alahly.com: 10 certificates, April 2021 to March 2022.
  • api-apple-buy.com: 10 certificates, February 2021 to January 2022.
  • eg-gov.org: 2 certificates, November 2021 to February 2022.

Three things fall out of that table.

A tight time window. The cluster is provisioned through 2021 and almost all of it expires in the first quarter of 2022. That is an operational lifecycle, infrastructure built for a campaign and allowed to lapse when the campaign ended. It lines up with the period when Predator operations against Egyptian and Greek targets were publicly documented.

A certificate-authority monoculture. Almost every certificate is from Let's Encrypt, the free, automated, 90-day authority. Free and automated is exactly what you want when you are spinning up disposable domains at volume. The one domain issued through a different cheap authority stands out as a candidate for a separate provisioning batch.

Deliberate isolation. None of the sampled domains share a certificate with another. Each stands alone on its own certificates. That is good operational security on the attacker's part, and it is also a finding: it tells you they compartmentalised, and that certificate-based pivoting will not collapse the whole set into one cluster. You note what the absence of a link tells you, not just the links.

Enrichment two: the registry, and the value of silence

The second passive source is domain registration data, through RDAP, the structured successor to WHOIS. Here the interesting result is a negative one. Query the campaign domains and most come back object not found: the registrations have lapsed and the records are gone.

That is the lifecycle again, seen from a second angle. The servers are down, the registrations are deleted, and if Certificate Transparency were not an append-only public log, the campaign would have left almost no trace at all. It is the reason CT is the first place you look for historical infrastructure: it is the one record the operator cannot delete after the fact.

Why this is the method, not a shortcut

Nothing above required contact with a single piece of attacker infrastructure. The certificate logs are run by third parties. The registry data is public. A journalist's safety was never put at risk by an investigator's curiosity. And yet the result is a dated, sourced reconstruction: what the campaign imitated, when it was built, how it was provisioned, how disciplined the operators were, and when they walked away.

At Ghost Protocol this is the same passive-first discipline behind DragonHunt, our OSINT and reconnaissance framework: establish what can be known from public records before anything active is considered, and keep a clean line between observing and interacting. For surveillance research that line is not just good practice, it is the whole ethic. The people on the other end of these domains were being hunted. The investigation has to be certain it never joins the hunt.

OSINT and reconnaissance, done cleanly

Ghost Protocol runs passive-first OSINT and reconnaissance from Colombo, for security assessments and threat research. If that is useful to your work, start with a conversation.

SOURCES // Predator domain indicators from the MVT project's Predator/Intellexa indicator set (569 domains), which gathers published research by several teams and indicators from Amnesty's Security Lab. Certificate history from public Certificate Transparency logs via crt.sh. Registration data via RDAP. All queries were passive, against third-party logs and registries, with no contact with any target infrastructure. Figures read on 2026-11-04 by Ghost Protocol (Private) Limited, Colombo. Imitated outlets and institutions are named only to show the lure pattern.

Get the next one

New writing on security engineering and AI memory, from work we actually did. How we handle your email.