Phishing, Vishing, Whaling, Smishing, and Watering Holes

A client wants to know whether employees will report a fake password-reset message. That sounds like a straightforward phishing test, but several decisions have to come first. Who may receive the message? What behavior is the client trying to measure? This episode explains how p…

A client wants to know whether employees will report a fake password-reset message. That sounds like a straightforward phishing test, but several decisions have to come first. Who may receive the message? What behavior is the client trying to measure?

What information may the test collect? What happens if someone enters a real password or contacts the help desk? Until those boundaries are clear, choosing a convincing message or sending platform is premature.

Social engineering reaches beyond a technical system. It affects people, business processes, support teams, privacy, and sometimes normal operations. A technically functional campaign can still be the wrong test if it contacts unauthorized people, collects unnecessary data, or creates consequences the client didn’t approve.

CompTIA PenTest+ objective 4.8 includes phishing, vishing, whaling, spear phishing, smishing, and watering-hole attacks. Those labels aren’t six interchangeable names for a deceptive message. You can distinguish them by communication channel, how narrowly the target is selected, and how the target encounters the interaction.

Phishing is the broad category most commonly associated with deceptive electronic messages. A campaign might imitate an account notice and encourage recipients to follow a link. Spear phishing is tailored to a particular person, role, or group.

The message may refer to the recipient’s job, project, or business relationship. The tailoring is the important difference. Sending the same generic lure to hundreds of addresses doesn’t become spear phishing merely because the sender possesses a recipient list.

Whaling narrows the targeting idea toward senior leaders or other high-value individuals. A message directed at a chief financial officer and crafted around an executive responsibility can be both whaling and spear phishing. The terms describe overlapping characteristics rather than mutually exclusive boxes.

Smishing changes the delivery channel to Short Message Service, or text messaging. Vishing uses a voice or telephone interaction. A deceptive account warning sent by text is smishing.

A caller pretending to be support and asking someone to take an action is vishing. The underlying pressure may resemble ordinary phishing, but the interaction changes the likely controls and evidence. An email filter can’t tell you whether an employee followed an instruction during a telephone call, while telephone records won’t explain what happened to an email message.

A watering-hole attack uses a different interaction model. Instead of directing a tailored message at each intended target, the attacker places the lure or malicious content somewhere members of the target population are expected to visit. The name comes from waiting at a shared destination rather than approaching each target individually. In an assessment, that distinction affects both authorization and interpretation.

Approval to send test messages to a named employee group doesn’t automatically authorize modifying a third-party website or exposing unrelated visitors. Likewise, evidence that a person visited an approved test page doesn’t prove that the person would disclose credentials. Channel, targeting, and interaction model tell you what kind of activity occurred. They don’t, by themselves, prove the impact or grant permission to perform it.

These distinctions also help you identify the control being evaluated. Security awareness and training may influence whether a person recognizes and reports a suspicious request. Access and authentication controls may limit what happens after a mistaken click.

Audit controls may record delivery, access, or reporting events. Those are related layers, but a result from one layer isn’t automatically evidence about another. If a mail gateway blocks every test message, you have evidence about message handling in that configuration.

You don’t have evidence that the intended employees would have recognized the lure. If employees report the message, that supports a conclusion about reporting behavior under the test conditions. It doesn’t establish that every technical control would block a real campaign or that the same employees would react identically to a voice call.

A useful test begins with a requirement stated as observable behavior. Consider the client’s request about fake password resets. The requirement could be to measure whether authorized recipients report the message through the approved reporting process.

That wording is more useful than saying the goal is to see who falls for phishing. It identifies a population, a behavior, and an evidence source. Before sending anything, predict what evidence a successful control should produce.

You might expect an authorized campaign record showing delivery and a corresponding record from the client’s reporting process. You might also need a defined way to distinguish delivery failures from messages that were delivered but never reported. The exact records depend on the approved environment, so they must be agreed upon rather than assumed.

Next, choose a bounded action that can produce the evidence without collecting more information than the objective requires. The message can point to a controlled page that records a harmless event, such as arrival at the page, and then ends the interaction. If the objective is reporting behavior, the test doesn’t need to ask for a production username and password. It certainly doesn’t need to retain one.

A client might want separate measurements for delivery, link selection, and reporting, but each measurement should be explicitly approved. Recording that an authorized test identifier reached a page is different from collecting the employee’s typed secrets, browsing data, or unrelated personal information. Realism isn’t a reason to make the collection unlimited.

The authorization should define the target population, timing, permitted channels, approved infrastructure, and data-handling rules. It should also account for dependencies. The support desk may receive calls.

The security operations team may detect and contain the campaign. A third-party mail provider may apply its own controls. None of those outcomes should be treated as an excuse to expand the test silently.

If the target list is ambiguous, clarification is the next action. If voice calls weren’t approved, an email exercise can’t be converted into vishing because delivery rates were poor. If a controlled test page would be reachable by the public, access and data collection need boundaries that protect people who aren’t participants.

Safe stopping conditions turn those boundaries into decisions during the campaign. Stop or escalate through the agreed channel if the activity reaches an unauthorized person, produces unexpected sensitive data, causes material operational disruption, or requires a stronger action than the client approved. The precise conditions belong in the engagement rules, not in an improvised judgment made after trouble appears.

This is especially important where operations have safety, reliability, or availability requirements. A social-engineering test aimed at people who support sensitive operations can create business effects even when the test payload is technically harmless. The lowest-impact method that establishes the required evidence should come before a more intrusive method.

After the bounded action, compare the records with the prediction. Suppose the campaign system reports that one hundred messages were submitted, the mail environment confirms delivery of eighty, and the reporting process records twenty reports. The submitted count alone can’t support a claim about one hundred employee decisions.

Twenty messages may not have reached their recipients, and the evidence doesn’t yet say why. Even among the eighty delivered messages, you’d need to know whether the reporting records can be matched to the authorized campaign without gathering unnecessary identity data. The defensible result describes what each record supports, states the denominator used, and preserves uncertainty rather than turning missing evidence into assumed behavior.

Conflicting evidence deserves troubleshooting before a stronger test. Imagine that the campaign platform shows a delivered message, but the recipient says it never appeared. Several explanations remain possible. A downstream control may have quarantined it.

The address may have been wrong or stale. The platform’s delivery status may represent acceptance by another system rather than appearance in an inbox. The test configuration or reporting integration may also be incomplete. None of these possibilities proves employee failure.

Check the authorized target data, configuration, expected control path, record timestamps, and data quality. The goal isn’t to force all records into agreement. It’s to determine which conclusion the available evidence actually supports.

The same care applies when the test page records a visit but the reporting system records nothing. A visit supports the narrow conclusion that the controlled resource received an approved interaction associated with the campaign’s measurement design. It doesn’t prove that a password was entered, an account was compromised, or the visitor understood the page. Automated security inspection might also reach links before a person does, depending on the environment.

The research provided for this episode doesn’t authorize a universal interpretation of every platform event. Validate the event against the expected mechanism and the client’s records. If you can’t distinguish a person’s action from a control’s action, report the condition as unresolved rather than assigning blame to an employee.

One especially risky assumption is that a realistic exercise must collect real passwords. It doesn’t. If the approved objective is to determine whether recipients report a password-reset message, a real credential adds risk without answering that question. Even when a client separately wants to measure whether someone attempts to submit information, a harmless substitute or controlled event can often establish the behavior.

The page might detect that an approved test field was used without requesting or preserving a production secret. The design must still be authorized and reviewed for privacy. The principle is to collect the minimum evidence needed for the stated test, not the maximum information the page can technically capture.

You can expose the bad assumption by asking what finding the real password would validate. Would possession of the password prove that the message was delivered? No. Would it prove that the recipient failed to report the message?

Not necessarily. Would it be needed to record that the person reached the test page? Again, no.

It would create a sensitive-data handling problem and possibly increase the consequences of an error. If the objective can be met through a safe marker or campaign event, collecting the actual secret is weaker engagement judgment. Technical capability doesn’t make additional collection necessary, proportionate, or authorized.

Troubleshooting should therefore separate a measurement problem from a security finding. If no reports appear, first verify that the approved reporting channel worked and that recipients knew how to use it under the test conditions. If no page events appear, check whether messages were delivered and whether the controlled destination operated as approved. If a result falls outside expectations, confirm target selection, prerequisites, timestamps, and record freshness before changing the campaign.

A more aggressive lure, a new communication channel, or broader targeting changes the test. Those changes require authorization, not merely frustration with incomplete evidence. Sometimes the correct result is that the available data can’t answer the client’s question.

Once validated, the results need context suitable for reporting. Describe the authorized population, delivery channel, targeting method, test period, expected behavior, and measurement limits. Separate observations from conclusions. An observation might state that the reporting process received a certain number of campaign-related reports during the approved period.

A conclusion might say that reporting behavior was observed among a defined portion of confirmed deliveries. If delivery confirmation was incomplete, state that limitation. If the organization’s controls prevented the message from reaching recipients, credit that control behavior without claiming the employees themselves recognized the message. Clear reporting allows the client to decide whether remediation belongs in awareness, message handling, reporting workflow, authentication protection, or some combination of controls.

The classification of the campaign belongs in that context. A generic email sent to the entire authorized population is phishing. A tailored message sent to payroll staff is spear phishing because it focuses on a role or group. A tailored message to a senior executive may also be described as whaling.

A text-message version is smishing, while a telephone interaction is vishing. A controlled destination placed where an authorized group is expected to visit follows the watering-hole model. Naming the technique correctly helps readers understand the tested channel and interaction, but the label isn’t the finding. The finding comes from validated behavior and evidence under the stated conditions.

Consider an organization that offers two possible tests. The first sends an approved fake account notice to a defined employee group and measures delivery and use of the reporting process. The second publishes a deceptive resource on an external community website because employees are believed to visit it. Both might produce interesting data, but the second option reaches beyond the organization’s direct population and infrastructure.

Without explicit permission from the site owner and protections for unrelated visitors, it conflicts with scope. The first option is the defensible choice when it meets the stated requirement with lower impact. This isn’t because phishing is always safer than a watering hole. It’s because this specific action has defined targets, approved infrastructure, and evidence aligned with the objective.

Now suppose the client has authorized text messages rather than email because the requirement concerns suspicious requests arriving on managed phones. The appropriate classification is smishing, and the evidence plan must follow that channel. Email delivery records would no longer establish whether the messages reached the managed phones.

If the client asks to replace the texts with live calls, that becomes vishing and changes the human interaction, timing, support burden, and evidence. The answer changes only after authorization and measurement rules change. Similar wording in the lure doesn’t make the channels equivalent.

Finally, think about the employee who responds in an unexpected way. The person might forward the message to a colleague, call a support number, disclose unrelated personal information, or report a genuine security concern outside the campaign’s intended path. The tester shouldn’t improvise a deeper interaction simply because an opportunity appears.

Follow the approved stop and escalation process, protect unexpectedly collected information, and document what occurred without overstating it. The engagement produces something more valuable than a tally of clicks. It produces evidence tied to a specific channel, target population, control behavior, and set of conditions, with enough qualification for the client to understand what was tested and what remains unknown.

Phishing terms become useful when they tell you how the target is reached, how narrowly the interaction is tailored, and which evidence should exist. Those distinctions let you separate a broad message from spear phishing or whaling, and distinguish email-like interactions from smishing, vishing, and a watering-hole model. The campaign itself still has to answer an approved question. Define the population, expected behavior, evidence, privacy limits, and stopping conditions before anyone is contacted.

Then use the least intrusive action that can establish the result. A realistic test doesn’t need real passwords, and a platform event isn’t automatically a validated finding. The defensible outcome is a carefully qualified conclusion that connects observed behavior to the authorized conditions under which it occurred.

Test your recall with this decision: What happens if someone enters a real password or contacts the help desk? The matching episode reference preserves exact terminology, named tools or standards, source-derived evidence checkpoints, and review material for Phishing, Vishing, Smishing, Whaling, Spear Phishing, and Watering Holes. This lesson was produced by Bare Metal Cyber dot com.

Phishing, Vishing, Whaling, Smishing, and Watering Holes

headphones Listen Anywhere

More Options »
Broadcast by