AI-FIRST CYBERSECURITY · Always Wise. Always Ahead. +91-95851-60363 info@semperwise.com DOVER, DE, USA

Attack Surface

They do the homework long before they touch your network.

Ask most teams when an attack starts and they describe something arriving — a scan, a phishing email, a failed login. Something in a log. That is the second phase. The first one is quieter, cheaper, usually more productive, and it never touches you at all.

SemperWise Threat Intelligence Team 6 min read Penetration Testing
New host exposed today MonWedFri Assets reachable from the internet

The opening move of a real attack leaves no trace on your side, because it is done entirely with public records. Most of them are records you published deliberately.

What is already public about you

Every TLS certificate you have ever issued

Certificate transparency is a public, append-only log. Every certificate any authority issued for your domain is in it, searchable, permanently — including the one for staging.internal.yourcompany.com that somebody spun up at 11pm two years ago. Nobody has to guess your subdomains. They can read the list.

Your DNS history

Passive DNS keeps records long after you change them: hosts you decommissioned, providers you migrated away from, and the valuable one — CNAMEs still pointing at cloud resources you no longer own. That last case is a subdomain takeover sitting there waiting for whoever checks first.

Your job adverts

A hiring post for a senior engineer is a technology inventory with a salary attached. Framework, cloud provider, CI system, monitoring stack, occasionally the exact versions. Written by you, published on purpose, indexed forever.

Your org chart

Professional networks make reporting structure legible by design: who runs finance, who joined three weeks ago, who is covering while somebody is on leave. A new joiner on a payments team is the most reliable phishing target in any company, and they are trivial to find.

Your code and your commit history

Public repositories carry more than code — internal hostnames in config examples, API endpoints in test fixtures, commit metadata, and occasionally a key that somebody removed in a later commit but not from the history, where it still lives.

Your email records

SPF, DKIM and DMARC are public DNS records. They state which providers may send as you, and whether a forged message from your domain gets rejected, quarantined or delivered to the customer. A permissive policy is an open invitation, published at your own request.

Why this is the phase that works

None of it requires a single packet aimed at your infrastructure. There is nothing to detect, nothing to rate-limit, nothing to raise an alert. It is fast, it is free, and — the uncomfortable part — it is largely lawful, because every source is one you or a public authority chose to publish.

By the time a scanner finally does touch you, it is not exploring. It is confirming a shortlist that was built while you were asleep.

It is almost never the main application

The pattern is consistent, and it is rarely the thing you have spent money hardening. The flagship product is tested, patched and watched. What gets found instead is the forgotten thing — a staging environment that was meant to be temporary, a vendor portal from a project that ended, a microsite for a 2023 campaign, a DNS record aimed at a storage bucket that was released years ago.

These all share one property: nobody owns them. So nobody patches them, and — this is the bit that matters — nobody remembers to put them in scope when the annual test is commissioned. They sit outside the test and inside the perimeter.

Run the same search on yourself this week

  1. Pull your own certificate transparency log.

    Search the public CT logs for your domain and read every single name that comes back. Expect surprises — everybody gets them. Anything you do not recognise is either yours and forgotten, or not yours at all, and both are worth an hour.

  2. Resolve every name you find.

    Whatever answers is live and in scope whether you meant it to be or not. Whatever points at a cloud resource you no longer control is a takeover risk, and it is the cheapest serious finding in this entire field.

  3. Read your own job adverts as an outsider.

    Describe the role and the problems, not the exact versions and vendors. You can hire excellent engineers without publishing a bill of materials.

  4. Check what your DMARC record actually says.

    If the policy is p=none, anyone can send mail as your domain right now and it will land. Moving to quarantine and then reject is roughly a week of work and closes an entire category of attack against your customers.

  5. Decommission properly, DNS included.

    Switching a server off is half the job. The record still pointing at where it used to be is the half that gets exploited.

  6. Then do it again next quarter.

    This is not a state, it is a feed. Certificates get issued, projects end, people leave, and the inventory starts drifting the week after you finish it.

The point

None of this is exotic tradecraft. It is a patient stranger with a browser and a free afternoon — and it is the standard opening of both a real attack and a competent assessment. The only difference between the two is which one gets there first.

Questions people ask about this

Is passive reconnaissance against a company illegal?
Reading public records is not. Certificate transparency logs, DNS, job adverts and professional profiles are published deliberately and intended to be read. The line is crossed when someone interacts with systems without authorisation — a scan, a login attempt, a request designed to provoke an error. That is exactly why the passive phase is so attractive to an attacker: it carries no legal exposure and no detection risk, and it narrows the target list before anything risky happens.
How do we find subdomains we have forgotten about?
Certificate transparency is the single best source, because a certificate had to be issued for the host to serve HTTPS at all, and that issuance is logged publicly and permanently. Passive DNS adds the hosts that never had a certificate. Between them you will usually recover names that no internal document lists. Resolve everything that comes back — the live ones are in scope for an attacker whether or not they are in scope for your test. Attack surface management covers doing this continuously.
How often should we review our external attack surface?
Quarterly at minimum, and continuously if you ship often or run more than one cloud account. The drift is not caused by attackers, it is caused by ordinary work: a new certificate, a contractor’s demo environment, a campaign site, a migration that left the old record in place. An annual review means up to twelve months of drift is invisible, which is longer than most forgotten assets survive unnoticed.

Sources & further reading

SemperWise Threat Intelligence Team

AI & Emerging Threats Research

Follows the frameworks, breach reports and adversary techniques shaping AI and agentic-system risk, and translates them into testing methodology and client guidance.

Next step

Talk to us about your next assessment.

A 30-minute call, then a scope document with what is in, what is out and what it costs. No obligation, and no charge for the conversation.