Social engineering is the whole family of attacks that trick a person into revealing information, granting access or moving money. Phishing is one technique inside that family: social engineering delivered by an electronic message, such as an email, a text or a chat message. MITRE ATT&CK also counts a voice call, as vishing. Every phishing attack is social engineering, but not every social engineering attack is phishing.
The difference matters because the two leave different traces. A phishing message carries a sender, a domain, a link or an attachment that filters can inspect. A call to your help desk, a flood of MFA prompts or a stranger at a badge door leaves nothing to scan, so a process has to catch it. Below, the two are compared side by side, nine real attack patterns are sorted, and each gets the control that stops it.
| Question | Social engineering | Phishing |
|---|---|---|
| What is it? | Any trick that gets a person to reveal information, grant access or commit fraud | Social engineering sent as an electronic message: email, text, chat or social media |
| How do they relate? | The whole category | One technique inside that category |
| Channels | Any: in person, phone, help desk, video call, a door, plus every phishing channel | Email, SMS, messaging apps, social media, calendar invites; ATT&CK also counts calls |
| What does it leave behind? | Often no file or link: ATT&CK says it works while minimizing technical indicators | A message, a sender domain and a link, attachment or number, often with a fake page |
| Targeting | Often a conversation that builds trust step by step, as CISA describes it | From mass spam campaigns to spearphishing of one person, company or industry |
| MITRE ATT&CK v20 | T1684 Social Engineering (Stealth), with Impersonation and Email Spoofing | T1566 Phishing (Initial Access) and T1598 Phishing for Information (Reconnaissance) |
| First line of defense | A rule to verify the requester through a channel the attacker does not control | Email authentication, filtering, phishing-resistant MFA and a report button |
What Is the Difference Between Social Engineering and Phishing?
The difference is scope. In NIST SP 800-63-4, social engineering means deceiving a person “into revealing sensitive information, obtaining unauthorized access, or committing fraud” by gaining their trust. NIST SP 800-115 calls phishing “a digital form of social engineering” that uses fake but authentic-looking emails to ask for information or lead users to a fake website. Both definitions sit in the NIST glossary.
CISA and MITRE draw the same line. CISA states it plainly: “Phishing is a form of social engineering.” MITRE ATT&CK is more precise: “All forms of phishing are electronically delivered social engineering.” Our social engineering glossary entry and phishing glossary entry define each term on its own.
Here is a one-question test: did the attacker deliver the trick in a message sent to the target, such as an email, text, chat or social media post? If yes, it is phishing. If it ran through a live conversation, a phone carrier, a door or an authentication prompt, it is social engineering of another kind. The one disputed case, a phone call, is covered below.
NIST’s own control catalog shows how wide the category is. Its control AT-2(3), in SP 800-53 Revision 5, covers training staff to recognize and report social engineering. The guidance lists phishing, pretexting, impersonation, baiting, quid pro quo, thread-jacking, social media exploitation and tailgating. Phishing is the first item on that list, not another name for it.
Is Phishing a Form of Social Engineering?
Yes. If a quiz asks “phishing is a form of social engineering: true or false?”, the answer is true. NIST, CISA and MITRE all place phishing inside social engineering. What phishing adds is a delivery method: a message, usually with a link, an attachment or a phone number to call.
The frameworks disagree at one edge, the phone call. Most of NIST’s phishing definitions name email and websites. MITRE ATT&CK counts a call as phishing, under Spearphishing Voice (T1566.004), while CISA describes vishing as “the social engineering approach that leverages voice communication.” In practice, log a call as vishing and defend it with the identity checks you use against other social engineering.
What Changed in MITRE ATT&CK in 2026
ATT&CK version 19, released on 28 April 2026, added Social Engineering (T1684) as a technique of its own. It sits under Stealth, one of two tactics created when that release split Defense Evasion. It has two sub-techniques: Impersonation (T1684.001) and Email Spoofing (T1684.002). The older Impersonation technique, T1656, was revoked and now redirects to T1684.001.
The T1684 description explains the split between the two terms. Social engineering gets users to take actions such as “password resets, MFA changes, financial approvals, or the disclosure of sensitive information,” while “minimizing technical indicators.” Phishing (T1566) stays an Initial Access technique, and Phishing for Information (T1598) a Reconnaissance one.
ATT&CK v20, released on 13 October 2026, is the current version and keeps this structure. Our T1566 phishing technique page covers the phishing side in more depth.
Which Is It? Nine Attacks Sorted
All nine patterns below are social engineering. The question is which of them are also phishing, because that decides which control can stop them. Each row comes from CISA, MITRE ATT&CK, the FBI, the UK NCSC or NIST. On a phone, scroll the table sideways.
| Attack and ATT&CK ID | Phishing? | What stops it |
|---|---|---|
| Mass email linking to a fake sign-in page (T1566) | Yes | Phishing-resistant MFA, link filtering |
| Text message with a link to a fake bank page (T1660, ATT&CK Mobile) | Yes (smishing) | Reporting the text; phishing-resistant MFA |
| Call to the IT help desk asking for a password or MFA reset (T1566.004, T1598.004) | Only in ATT&CK (vishing) | Verifying the caller before any reset |
| Repeated MFA push prompts until the user taps Accept (T1621) | No | Number matching, then phishing-resistant MFA |
| Persuading a mobile carrier to move a number to the attacker’s SIM (T1451, ATT&CK Mobile) | No | MFA that does not rely on SMS or voice codes |
| Email from a supplier’s hijacked mailbox asking to change bank details (T1684.001) | Yes, when it arrives by email | Confirming account changes through a second channel |
| Fake error asking the user to paste a command, known as ClickFix (T1204.004) | Often, as the lure | Application control, PowerShell Constrained Language mode |
| Voice or video call with a cloned executive asking for a transfer (T1684.001) | No, unless the invite came by phishing | Hanging up and calling back on a known number |
| Following an employee through a badge door, or tailgating (physical, not mapped) | No | Training staff to spot and report it |
Read the table by its last column. Rows that are phishing can be caught partly by technology, because there is a message, a domain and a page to inspect. The other rows run through a person who can approve something, so the control is a rule for how that approval gets checked.
Scattered Spider shows both halves in one group. CISA’s advisory AA23-320A says its intrusions historically began with broad phishing and smishing, using domains such as targetsname-sso[.]com and targetsname-helpdesk[.]com. The group then moved to targeted spearphishing and vishing. It calls help desks to reset passwords or move MFA tokens, and it uses SIM swaps and push bombing.
Phishing: The Part You Can Inspect
ATT&CK names many delivery routes for phishing: attachments, links, third-party services such as social media, voice calls and even calendar invites. Spear phishing targets one person, company or industry. Smishing uses text messages, and vishing uses calls. The Phishing 101 guide walks through the warning signs.
Every phishing message leaves evidence: a sender, a domain, a link or a phone number. That is why technical controls work against it. The catch is that sender checks prove which domain sent a message, not whether that domain is honest. In this illustrative header, a message from a lookalike domain passes all three checks:
Authentication-Results: mx.example.com;
spf=pass smtp.mailfrom=examp1e-billing.example;
dkim=pass header.d=examp1e-billing.example;
dmarc=pass header.from=examp1e-billing.example
From: "Example Corp Accounts" <invoices@examp1e-billing.example>
; Your own domain: a DMARC policy record at enforcement (RFC 9989)
_dmarc.example.com. IN TXT "v=DMARC1; p=reject; rua=mailto:dmarc-feedback@example.com"
The checks pass because the attacker owns examp1e-billing.example and set it up properly. RFC 9989, the DMARC standard published in May 2026, says DMARC “does not address the use of visually similar domain names” or a fake display name. DMARC on your own domain still matters. At p=reject, it tells receivers that mail failing the check is not a valid use of your domain.
Lookalike domains need a different control: watching new registrations and taking down the ones that copy you. Brandefense phishing monitoring finds lookalike domains and fake pages that copy your brand and gets them taken down. Our lookalike domain response guide lists the steps.
Social Engineering That Never Sends a Phishing Message
The rest of the category leaves little for a filter to find. ATT&CK’s T1684 description names executive, vendor and help desk scenarios, including AI-enabled voice calls, aimed at password resets, MFA changes and payment approvals. The request looks like routine work, which is the point. Four patterns stand out in the advisories cited here.
The Help Desk Call
CISA describes Scattered Spider calling employees and help desks over several calls, first to learn the reset steps and then to get passwords reset or MFA tokens moved. The callers come prepared with personal details from social media, open sources and leaked databases. In May 2025, after attacks on retailers, the UK NCSC told organizations to review how their help desk authenticates staff before a reset, especially for privileged accounts.
MFA Push Bombing and SIM Swaps
Push bombing, also called MFA fatigue, floods a user with sign-in prompts until one is approved. CISA counts it among Scattered Spider’s social engineering methods. The user receives prompts from a real system, not a phishing message.
A SIM swap targets the mobile carrier instead of the user. CISA’s MFA fact sheet calls it “a form of social engineering” in which attackers convince carriers to move a phone number to a SIM card they control. SMS codes for that number then reach the attacker.
Impersonation by Email, Phone or Video
Business email compromise is impersonation aimed at money. The FBI describes BEC as fraud carried out “through social engineering or computer intrusion techniques” by compromising email accounts, phone numbers or virtual meeting apps. Its 2025 report adds that voice cloning can be used to request wire payments, and businesses reported over $30 million in losses to BEC involving AI. Our business email compromise entry and deepfake and voice clone fraud guide cover both sides.
Tricks That Start Away From the Keyboard
NIST’s list includes tailgating and baiting next to phishing. CISA’s own description of a social engineering attack is a person, not a message. It is someone claiming to be “a new employee, repair person, or researcher” who asks questions until they can piece together enough information to get into the network.
Pretexting, Baiting, Quid Pro Quo and Scareware
Security courses give the non-phishing tricks their own names. Pretexting is an invented story or role used to get information, like the help desk caller above. Baiting is a lure, such as a free download or a found USB stick. Quid pro quo offers a favor, such as fake IT support, in exchange for access, and scareware shows fake alerts that push a download or a payment.
They work for the same reason phishing works: they pressure a person to act before checking. CISA notes that an attacker “may use a false sense of urgency or importance.” It also describes an attacker who cannot learn enough from one person and asks another, relying on the first answer “to add to his or her credibility.”
Where the Two Meet: What Phishing Kit Ads Show
The line between the terms blurs inside real phishing products. For the Fraud chapter of our Cybercrime Report 2026, Brandefense compared eight full phishing advertisements saved in Brandefense Blackhole, our cybercrime database. These are sellers’ claims, not tested tools or confirmed attacks.
A 16 May 2026 advertisement promised to bypass every form of two-factor authentication. The same text then said the system would wait for the user’s approval or SMS code. So a product sold as a technical bypass still relies on a person’s response. Our analysis of what phishing offers target beyond a password covers all eight ads.
The same ads show why a channel label is not enough. A June panel listed SMS, calls and WhatsApp among its features, and an October dashboard listed an SMS sender next to a device-code login feature. A text or a call says how a message reached a person. It does not say whether the attacker wanted a password, a session or an approval.
ENISA sees the same overlap at scale. Its Threat Landscape 2026 puts phishing at 77.8% of the social engineering techniques seen in incidents it identified against the EU, followed by malicious spam at 13%. ENISA adds that device-code phishing and ClickFix show a shift toward abusing trust “rather than technical weaknesses alone.” Our summary of ENISA’s phishing findings has the details.
Which Costs More: Phishing or Other Social Engineering?
The FBI’s 2025 Internet Crime Report shows a gap between how often each attack is reported and what it costs. Phishing and spoofing was the most reported crime type, with 191,561 complaints and $215.8 million in reported losses. Business email compromise drew far fewer complaints, 24,768, but $3.05 billion in losses.
Impersonation scams sit near the top as well. Complaints about fraudsters posing as tech or customer support reported $2.13 billion in losses, and government impersonation $797.9 million. These are complaints filed with the FBI, so they show what victims reported, not every incident. Still, the pattern is clear: phishing is the most common contact, and the largest losses follow when a person is persuaded to pay or approve.
How Do You Defend Against Each One?
The split in the comparison table becomes two sets of controls. Phishing controls inspect the message and harden the sign-in. Controls for other social engineering add a verification rule wherever one person can approve something on another person’s word.
Controls That Stop Phishing
- Phishing-resistant MFA. CISA calls FIDO/WebAuthn “the only widely available phishing-resistant authentication.” Its Scattered Spider advisory adds that it is not susceptible to push bombing or SIM swaps. CISA’s fact sheet shows why app codes are not enough. In a common phishing technique, the user types a password and the 6-digit authenticator code into a site that mimics the real login portal. With FIDO/WebAuthn there is no code for a caller to ask for and no push to approve by mistake.
- Number matching as a stopgap. Where push MFA stays for now, CISA recommends number matching as an interim defense against MFA fatigue.
- DMARC at enforcement. Publish p=reject on every domain you send from, with aggregate reports switched on, so exact-domain spoofing fails.
- Lookalike domain watch and takedown. DMARC does not cover visually similar domains, so someone has to find and remove them.
- A clear way to report. CISA’s advice is to report the message, then delete it without clicking any link.
Controls That Stop Other Social Engineering
- A reset rule for the help desk. No password or MFA reset on an inbound call alone. Call back on the number in the employee record, and add a second approver for privileged accounts.
- A payment change rule. The FBI’s IC3 advises using secondary channels to verify requests to change account details. Use a number you already hold, never one from the request.
- A call-back habit for voice and video. For calls that may use a cloned voice, the FBI’s advice is to hang up, look up the organization’s number yourself and call it directly.
- Training that covers more than email. NIST AT-2(3) covers recognizing and reporting every form of social engineering, including pretexting and tailgating, not only phishing tests.
How Should a Security Team Label the Incident?
Calling every case “phishing” hides the step that needs fixing. If a help desk call led to the breach, a better email filter would not have stopped it. Our research on phishing ads makes the same point: record the channel, the service imitated, what was requested or approved, and the access actually observed as separate facts.
A short incident record keeps those facts apart. The fields below are an illustrative template, not a standard, but the ATT&CK IDs are real:
incident: help desk MFA reset
channel: phone call to the IT service desk # vishing
pretext: caller claimed to be a named employee # impersonation
preceded_by: none found # or an SMS with a link
request: register a new MFA device
approved_by: service desk analyst, no call-back
access_observed: new MFA device, SSO sign-in from a new IP
attack_ids: [T1566.004, T1684.001]
control_gap: reset allowed on an inbound call alone
For detection, ATT&CK’s Detect Social Engineering strategy (DET0899) offers a useful rule of thumb. One of its analytics looks for consent grants, password resets, role changes, external sharing or token creation shortly after a user deals with messages, invites or help desk workflows. Joining a sensitive change to a recent outside contact can surface both phishing and the calls that never sent a message.
What to Check This Week
- Ask your help desk how it verifies a caller before a password or MFA reset, then test it with a planned call.
- Look up the DMARC record of every domain you send mail from, and plan the move to p=reject.
- List the accounts that still accept SMS codes or simple push approvals, starting with admins and finance.
- Write the payment change rule down and make sure finance applies it to every supplier.
- Find out who is registering lookalike domains of your brand, and have a takedown route ready before you need it.
Frequently Asked Questions
Is phishing a form of social engineering?
Yes. NIST, CISA and MITRE ATT&CK all treat phishing as a form of social engineering. MITRE puts it this way: "All forms of phishing are electronically delivered social engineering." So the statement "phishing is a form of social engineering" is true.
What is the main difference between social engineering and phishing?
Scope and delivery. Social engineering is any trick that gets a person to reveal information, grant access or commit fraud, in person, by phone or online. Phishing is the part delivered by an electronic message, such as an email, text or chat message, usually with a link, an attachment or a number to call.
Is vishing phishing or social engineering?
Both, depending on the framework. CISA calls vishing a social engineering approach that uses voice. MITRE ATT&CK files it under phishing as Spearphishing Voice (T1566.004). Whatever the label, defend calls with identity checks, such as calling back on a number you already hold.
Is business email compromise phishing?
BEC is social engineering aimed at money. When the request arrives by email, it is also phishing, and ATT&CK maps the impersonation to T1684.001. The FBI notes that BEC also runs through phone numbers and virtual meeting apps, so confirm every payment change through a second channel.
Which is more dangerous, phishing or social engineering?
In the FBI's 2025 Internet Crime Report, phishing and spoofing was the most reported crime type, with 191,561 complaints and $215.8 million in losses. Business email compromise had 24,768 complaints but $3.05 billion in losses. Phishing is the most common contact, and the biggest losses follow when someone is persuaded to pay.
Which MITRE ATT&CK technique covers social engineering?
Since ATT&CK v19, released in April 2026, Social Engineering is technique T1684 under the Stealth tactic. Its sub-techniques are Impersonation (T1684.001) and Email Spoofing (T1684.002). Phishing remains T1566 under Initial Access, and Phishing for Information is T1598 under Reconnaissance.
