✦ Preferences saved
TRACE / ANAF + ANCPI / PUBLIC SECURITY / FOLLOW-UP
24 APRIL → 14 JULY → 19 JULY → 01 OCTOBER 2026

THEY WERE WARNED. THE ATTACK FOLLOWED. I CHECKED AGAIN.

ANAF + ANCPI. From the record I published in April, to the ANCPI ransomware incident in July, then the technical re-check in October.

In April I published the ANAF technical record. In July, ANCPI was hit by ransomware. In October I returned to the public ANAF and ANCPI surface to check WHAT CHANGED and WHAT REMAINED. Below I separate what I observed, what the evidence demonstrates, and what the institutions operating these services still need to explain or remediate.

OPEN THE RECORD ↓SEE WHAT I FOUND →
01 / THE RECORD

A confirmed incident. Then a new public snapshot.

On 14 July 2026, ANCPI detected unauthorized access. The Romanian Government later stated that the technical investigation confirmed ransomware and that attackers encrypted and deleted part of the virtualization infrastructure hosting agency applications. The October audit is a separate event: it records what public services exposed at the time of capture.

The October observations do NOT prove that they caused the July ransomware incident. They show current public controls that can be checked independently.
JULY 2026

Confirmed ANCPI ransomware incident, service disruption, investigation and recovery.

OCTOBER 2026

I returned to the public-facing ANAF and ANCPI services to see what had changed and what remained. Below, I document each observation separately and the public material it is based on.

02 / WHAT THE AUDIT FOUND

Concrete deficiencies, separated from speculation.

ANCPI HTTP remains usable

The captured request to http://ancpi.ro/ ended on HTTP and returned content. The observed public configuration did not force that request onto HTTPS.

REAL CONTROL RISK: traffic delivered over HTTP lacks TLS confidentiality, integrity and server authentication.

ANCPI DMARC is monitoring-only

The published DMARC policy was p=none. That requests reporting, not quarantine or rejection based on DMARC failure.

REAL CONTROL RISK: the domain does not request DMARC enforcement. Actual mail acceptance still depends on SPF, DKIM, alignment and receiver policy.

ANAF JSESSIONID lacks browser protections

Captured ANAF responses emitted JSESSIONID without HttpOnly, Secure and SameSite attributes. The observation was repeated across multiple captured responses.

REAL CONTROL RISK: those explicit browser-side restrictions are absent. The evidence does not prove XSS, CSRF, session theft or the role of that cookie.

ANAF public hardening gaps persist

Captured responses also showed missing HSTS and other browser-side hardening controls, plus public middleware fingerprinting. These are configuration findings, not proof of a successful attack.

REAL CONTROL RISK: missing hardening removes defensive layers and public fingerprinting reduces uncertainty for an attacker.

03 / REPRODUCIBLE RUN

See the evidence as a run, not as a slogan.

This terminal is a deterministic presentation of the preserved audit snapshot. It does not contact ANAF or ANCPI from your browser.

$ trace-evidence --snapshot 2026-10-01
ready · sanitized evidence snapshot loaded
04 / EVIDENCE CHAIN

CLAIM→EVIDENCE→CONSEQUENCE→WHERE I STOPPED.

Believe me or don't. That's your business. I bring the evidence. For every important point here I show what I found, when I found it and what follows concretely from it. I stopped where I decided to stop, not because that was necessarily as far as I could have gone. Everything I publish can be checked independently. The files have SHA-256 identifiers so you can verify that what you download is exactly what I published. The hash proves the integrity of the file. That's it. By itself it does not prove the cause of an incident and it does not turn an assumption into fact.

05 / CONTINUITY

IN APRIL I PUBLISHED WHAT I FOUND. IN OCTOBER I WENT BACK.

In April I made public what I found in ANAF's infrastructure: old technology, errors exposed directly to users, and a system raising enough questions to be worth documenting. I am not deleting anything, and I am not rewriting the past after what happened in the meantime. The April dossier stays there, exactly in the context in which I published it. In October I went back. This time I checked the public-facing ANAF and ANCPI surface separately and preserved the evidence again. Two different moments, two different captures. I connect them through chronology; I do not mix them together just to make a story fit.

OPEN THE APRIL DOSSIER →
06 / AFTER THE RANSOMWARE

AFTER THE RANSOMWARE, QUESTIONS REMAIN THAT SOMEONE NEEDS TO ANSWER.

After the ransomware, public discussion focused on recovery, safety and remediation, but the October capture still shows concrete things that call for an explanation: ANCPI's public root was reachable over HTTP without forced HTTPS, DMARC was still set to p=none, and captured ANAF responses were missing hardening controls that I documented separately. I am not saying the ransomware got in through any of these things, because I do not have evidence proving that and I do not need to invent it. I say what I found, I publish what I can demonstrate, and I ask why these things were still there.

The questions are simple, and the evidence is public.

Why was ANCPI's public root reachable over HTTP without forced HTTPS in the October capture?
Why was DMARC still set to p=none, and what is the plan for moving to enforcement?
Why did captured ANAF responses issue JSESSIONID without Secure, HttpOnly and SameSite?
Which hardening gaps in the published capture have been remediated since, and when?
07 / SOURCES & BOUNDARIES

BEFORE ANY ARGUMENT, THESE ARE THE SOURCES I BASE WHAT I SAY ON.

I do not build conclusions from hearsay, and I have no intention of dressing up what the evidence shows. Romanian public institutions are not facing their first security incident, and ANCPI cannot be treated as an accident that appeared out of nowhere. Critical vulnerabilities had been reported years earlier, and after the ransomware attack public hearings described old and unpatched systems, weak and reused passwords including administrator accounts, and permissive internal-network access rules. I call that institutional incompetence, and I stand by that wording publicly. Conditions like these should not have been tolerated in public infrastructure of this importance, especially after warnings and repeated incidents. I checked from the outside, as a self-taught researcher, out of passion and on my own time. I publish here what I can demonstrate up to this point. That does not mean these are all the things that could be discovered, or that this is the furthest anyone could go. In the public audit, I stopped where I decided to stop. At the same time, I will not turn suspicion into evidence just to make the text sound more dramatic: my October findings do not by themselves prove how the July ransomware entered. For what I state, there are sources, captures and evidence. For what I cannot yet demonstrate, I do not invent.

THE TEST IS SIMPLE

I HAVE DONE MY PART. WHAT REMAINS ARE THE EVIDENCE AND THE QUESTIONS.

State institutions have enormous budgets, contracts, specialists, entire structures and responsibilities to match. When that ends with systems left old and unpatched for years, reused administrator passwords, vulnerabilities reported in advance and, ultimately, infrastructure erased by ransomware, I call the result amateurism and institutional incompetence. I have no reason to dress it up. I did not have a public budget, sponsors or a paid team behind me. I had my own computer, my own time, the knowledge I built for myself, curiosity, patience and the persistence to keep checking until the pieces came together. Even so, I found enough for the dossier you see here to exist. I publish what I can demonstrate up to this point. I stopped where I decided to stop. That does not mean this is where everything that can be found ends, nor where further verification has to stop. What makes it more absurd is that many of these things can be done better. I have plenty of ideas for improving the systems, controls and the way infrastructure like this is administered. It is not a lack of possibilities that angers me. It is the gap between the resources available to these institutions and the results we can see.

The evidence is public. Anyone who thinks I am wrong can start exactly there.

TRACE