Cybersecurity for small business: what an attacker sees from the outside, and what to fix first
Small business cybersecurity starts with knowing what anyone on the internet can see about your company. Seven checks, most of them with free public tools: what the results mean and what to fix first.
In 30 seconds
- What happens
- Seven checks, most with free public tools, show what anyone on the internet can see about your company, plus an order for fixing what you find.
- Who it affects
- Small businesses with a website, company email or remote access, especially when an agency or outside provider manages them.
- First step
- Gather your company's domains and search crt.sh for subdomains you do not recognize, then find out who manages each one.
- Our view
- Before buying another tool, a small business should know what it exposes to the internet. An exposed system can reveal outdated software or an open login page, and often no one is in charge of it. A list of what is exposed, with an owner for each system, lets you decide what to close and what to update. Without that list, a new tool is set up only on the systems you remember.
Start by finding out which of your company's sites, logins and data can be found from the internet. You can run several of the seven checks below yourself with public tools. For others you will need whoever manages your website or your email. Have a list of your domains ready, along with that person's contact details.
Verizon's 2026 DBIR, an annual data breach report published in May 2026, breaks down how attackers got into the breaches of small organizations (fewer than 1,000 employees). Exploiting software flaws (vulnerabilities) was the way in for 26% of them. Stolen or leaked passwords account for 13%, and phishing (fake emails that trick people) for 9%. Systems and login pages open to the internet are one of the first places to look for the first two.
Seven things an attacker sees from the outside
Run these checks only on your own company's domains and systems. Some look up public information and others use your own admin panels. None involves attacking a system. You will need the list of the company's domains (the one for your website, the one for your email, and any from old campaigns or brands) and, if you know them, your public IP addresses. Checks 2 and 6 are the most technical, so you may want to do them with whoever manages your systems.
1. Forgotten subdomains and sites
When a website gets a security certificate (the thing behind the padlock in the browser), that certificate is recorded in public logs called Certificate Transparency logs. Subdomains the company has forgotten, names like promo.yourdomain.com, can show up there: for example, the site of a campaign that ended years ago.
How to check. On crt.sh, a free search engine for those logs, type %.yourdomain.com (with your own domain). You will see the names certificates were issued for. Some will be repeated or old, and an old certificate does not prove the site is still online. What matters are the names you do not recognize.
If you do not recognize a name, find out who manages it before you conclude it is abandoned. A system with no owner is a system nobody updates. In one of our cases, forgotten servers still had factory passwords, and some test environments still held real data.
What to do. For each name, confirm whether it is still online and who manages it. Take down what is no longer needed and put the rest on the list, with its owner. PHP2Shell and unknown PHP sites explains the risk of leaving a site off the inventory.
2. Open admin panels and remote access
Remote Desktop (RDP, a way to control an office computer from elsewhere), VPN login pages (the VPN is the remote connection into the office network), the management panels of a firewall or router, cameras and printers can end up reachable from the internet through a setting nobody reviewed.
How to check. Shodan is a search engine that crawls the internet and records which services answer at each IP address. With an account you can search hostname:yourdomain.com. Censys, a similar search engine, offers a limited number of free searches each month; the amount can change. If you do not know your public IP address, ask your internet provider. What these search engines show can be up to 30 days old, so confirm it with whoever manages the devices.
The management panel of a firewall, a router or a server should not be open to the whole internet: firewall and router vendors commonly recommend limiting it to trusted addresses. In another of our cases, a VPN router answered management queries over SNMP, a network administration protocol, and gave away usernames and device details. Our article on a chain of flaws in PAN-OS firewall software shows what can happen when a vulnerability turns up in an exposed panel.
What to do. Block access from the internet to any service that does not need to be reachable from outside. For Remote Desktop and management panels, ask your provider for controlled access with two-step verification instead of leaving them exposed, and keep the built-in software (firmware) current on the devices that do have to be reachable.
3. Outdated software on the website
Your website runs on a content management system (WordPress, Joomla, PrestaShop), its plugins and themes, and a version of PHP, the programming language many websites run on. An old version may have flaws that are already public, and attackers look for them automatically.
How to check. Log in to the site's admin panel: WordPress, for example, shows pending updates for WordPress itself, the plugins and the themes. Look up the PHP version in your web hosting panel or ask the provider.
PHP versions that are out of support no longer get official patches, and neither does a plugin its author stopped maintaining. Flaws found later will stay unfixed for as long as you keep using that version.
What to do. Update what you use and remove what you do not. If an agency maintains the site, ask them in writing which versions are installed and how often they update them.
4. Work passwords in leaks
When an outside service suffers a breach, its users' email addresses can be exposed and, sometimes, their passwords. If an employee signed up with a work address and reused the password, the company account is at risk even though the company itself was never breached.
How to check. On Have I Been Pwned you can check one email address for free, without signing up. The result says which breaches it appears in and what kind of data each one included. A search of the whole domain requires proof that the domain is yours, and it is free for domains with up to 10 breached addresses.
An address in a breach does not prove anyone got in. The real risk is when the breach included passwords and the same password is used for work email or the VPN. Something like that happened in our case of a work password exposed through an outside service. Another route is malware that steals passwords saved in the browser, which we explain in what an infostealer is.
What to do. Change the passwords that may have been reused, turn on two-step verification on those accounts and agree on a simple rule: the work address is not used to sign up for personal services. If there are signs that someone got in, also sign the account out on all devices and check the mailbox for forwarding rules you did not create.
5. Domains that look like yours
Someone can register a domain with one letter changed, another ending (.net instead of .com) or an added hyphen. Such domains can be used to send email that looks like yours to customers and suppliers.
How to check. Free tools such as dnstwist generate the variants of your domain and show which ones are registered.
A variant registered by someone else and set up to receive email deserves attention, though it does not prove anyone intends to impersonate your company. We explain this in lookalike domains and typosquatting.
What to do. Remind your team and your key customers which domain is the real one, and talk to your registrar or a lawyer if the lookalike uses your brand.
6. Email that can be faked
Three mechanisms, published in your domain's DNS (its public settings), help receiving servers tell your email from a fake. SPF says which servers may send on your behalf. DKIM adds a signature to each message that is checked against a published key. DMARC says what to do with messages that fail those checks.
How to check. Check MX, in Google Admin Toolbox, reviews your domain's mail and SPF records, and MXToolbox's DMARC lookup shows your DMARC policy. DKIM needs the name of the key (the “selector”): ask your email provider whether it is turned on, or send yourself a message, open its details and see whether it passed the DKIM check.
Without DMARC, or with DMARC in monitoring-only mode (p=none), your domain does not ask receiving servers to reject a fake email that uses your exact address. Also, since February 2024 Gmail has required all three from anyone who sends more than 5,000 messages a day to Gmail accounts (Google allows p=none for this).
What to do. Ask whoever manages your email to publish SPF and DKIM and to move DMARC, in stages, to p=quarantine or p=reject once legitimate mail is confirmed to pass the checks.
7. What the company publishes about itself
A company publishes more than it thinks: email addresses on the website, an org chart anyone can piece together from LinkedIn, vendors named in case studies, technologies listed in job postings.
How to check. Search for the company's name the way a stranger would: its website, its LinkedIn profiles and its job postings.
Publishing this information is not a security flaw in itself, but it is the material a convincing phishing email is built from: who pays the invoices, which bank or supplier the company works with, who just joined.
What to do. Use generic addresses (billing@, info@) instead of personal ones where a name is not needed, and agree that no change of bank account or urgent payment is approved by email alone. Confirm it by phone, using a number you already have on file.
Write down the date of each check, because results change whenever a new site goes live or someone opens up remote access. A platform that watches what your company exposes to the internet (attack surface monitoring) repeats some of these checks on a schedule: exposed systems, outdated software it recognizes, leaked passwords. It puts the results in one place. Your own admin panels and what the company publishes about itself still need a person to check.
Small business cyber security checklist: what to fix first
If you find more problems than your team can fix this week, here is a practical sequence to work through. If there are signs that someone is already in, or a serious finding on an exposed system, that comes before everything else.
- Two-step verification on email and remote access. CISA, the US cybersecurity agency, recommends in its Cyber Essentials guide (unrelated to the UK certification of the same name) starting with people who have administrator or remote access. Email belongs in the same group: with two-step verification, a leaked password on its own is usually no longer enough to get in.
- Close or update what is open without a reason. Management panels, Remote Desktop, forgotten sites: take anything that does not need to be online off the internet, and update the rest.
- A list of what is out there, with an owner for each system. Domains, sites, remote access, cloud services. A spreadsheet is enough to start.
- Isolated backups. At least one copy disconnected from the company network, under its own protected account, and a restore test scheduled on the calendar.
- A short phishing briefing. How to spot a suspicious email and, above all, whom to tell without fear of looking foolish.
What a one-off review does not see
What the company exposes changes. Someone publishes a new site, a vendor opens remote access to provide support, a vulnerability is published in a device that had no known flaws until then. A review done today does not cover the changes that come after it. More vulnerabilities, less time explains why the gap between reviews matters more each year. Free vulnerability scanner or perimeter monitoring covers the difference between a single review and continuous monitoring.
Repeat the checks after every major change and when a security advisory is published for a device or software you use. If you automate the external checks, make someone responsible for reading the alerts.
When to ask for help, and what to ask for
Ask for a specialist review if you find systems nobody can account for, if the company has an online store or a customer portal, or if a customer, an insurer or a regulator asks for evidence of your security. After an incident, get help to respond and to investigate what happened.
Before you hire anyone, be clear about what you need checked. A vulnerability scan looks for known flaws in the systems you agree on. A penetration test checks, within an agreed scope and conditions, whether someone could get in and how far they would get. We cover the difference in vulnerability scan or penetration test: what each one answers. Before you order a test, also read what a penetration test report should include.
Where to turn depends on the country. United States: CISA, the federal cybersecurity agency, takes incident reports, and the FBI's Internet Crime Complaint Center (IC3) takes complaints about cybercrime and fraud. United Kingdom: Report Fraud, the service that replaced Action Fraud, takes reports of cybercrime and fraud from England, Wales and Northern Ireland, and has a phone line for businesses under a cyber attack (0300 123 2040). In Scotland, reports go through 101. The NCSC's page “Where to report a cyber incident”, on GOV.UK, points to the right body for each case. Spain: INCIBE's 017 helpline is a free service that advises companies and individuals. Other EU countries: ask the national incident response team (CSIRT). The CSIRTs Network site has a map of the teams by member state. If a crime such as fraud is involved, report it to the police as well.
Sources
- Verizon, 2026 Data Breach Investigations Report, May 2026, section on small and medium-sized businesses.
- CISA, Cyber Essentials, 2019 guide, accessed November 13, 2026.