Help desk and IT support
MGM Resorts (2023)
One phone call to the IT help desk, pretending to be an MGM employee, was enough. Scattered Spider got MFA reset, took over identity systems, and set off an outage.
Case library
Social engineering examples are real attacks where criminals trick people into handing over access, data, or money, usually by pretending to be someone the victim trusts. This library documents 172 of them. Every case names who was targeted, how the attacker got in, what it cost, and where the facts come from.
Every one of these is a documented case, and its sources are on the case page. They’re grouped by the trick that got the attacker in, since that’s the part a defender can actually change.
Help desk and IT support
One phone call to the IT help desk, pretending to be an MGM employee, was enough. Scattered Spider got MFA reset, took over identity systems, and set off an outage.
Help desk and IT support
Attackers, later tied to Scattered Spider, talked their way past Caesars’ outsourced IT support.
Help desk and IT support
Attackers rang Twitter staff pretending to be the IT help desk and collected VPN logins. Then they used internal admin tools to hijack 130 high-profile accounts for a "double your bitcoin" scam.
Help desk and IT support
An attacker had a contractor’s stolen password. They flooded the contractor with MFA push prompts, then posed as tech support to get one approved.
Executive or supplier
A finance worker in Arup’s Hong Kong office joined a video call where, according to Hong Kong Police, everyone else on screen was a deepfake. The employee then made 15 transfers.
Executive or supplier
Fraudsters posing as Ubiquiti’s CEO and a partner at an outside law firm got its principal financial officer and controller to wire $46.7M from the Hong Kong subsidiary’s account, in 14 transfers over two weeks.
Executive or supplier
A Lithuanian fraud ring pretended to be a real hardware supplier, Quanta Computer. With spoofed emails and forged invoices, it got both companies to pay into accounts it controlled.
Phishing email
It started with a phishing email at a small HVAC contractor. That stole the contractor’s login for Target’s vendor systems, and the attackers used it to reach Target’s network.
Phishing email
A spear-phishing email titled "2011 Recruitment Plan," carried a booby-trapped Excel file. It got attackers into RSA and exposed SecurID data.
Phishing email
North Korean operators sent Sony staff fake Apple ID "verify your account" emails and collected reused passwords. Later they set off wiper malware that crippled the studio.
Text message
A text, then a phone call from a fake IT colleague, was all it took for a Retool employee to hand over one-time codes.
Fake job candidate
KnowBe4 hired a North Korean operative as a software engineer. He’d passed four video interviews, using an AI-enhanced stock photo.
Physical baiting
An infected USB flash drive went into a U.S. military laptop, and the agent.btz worm spread from there into classified networks.
146 full case pages are published so far, and the rest follow in stages. Linked names open the full case.
The ten main types are phishing, vishing, smishing, quishing, pretexting, physical social engineering, deepfakes, help-desk and MFA manipulation, ClickFix and SEO poisoning, and attacks run by AI agents. Below, they’re ranked by how often each one turns up in real cases. Some attacks crossed channels, so the counts overlap.
An email that pretends to come from someone you trust, to steal a password or get a payment sent.
A phone call from someone posing as IT, your bank, or the boss, who talks you out of your doubts in real time.
Getting into a building, or planting a bad device, by following an employee through the door or posing as a contractor.
A made-up but believable story that explains why the attacker is asking, set up before any request comes.
AI that researches the target, writes the lure, and keeps the conversation going, so one person can run lots of tailored attacks at once.
A cloned voice or face on a call or video, so you think you’re talking to someone you know.
The same con by text or messaging app, where there’s no sender address to check.
A QR code hiding the bad link, which moves the click onto your own phone, outside company filters.
A rigged search result or fake error message tells you to paste in a command, and you end up installing the malware yourself.
Calling the service desk as a locked-out employee to get MFA reset. Account recovery becomes the way in.
Ranked by how many cases in this library involve each one. Some attacks used more than one channel, so they overlap.
| Name | What it means | Closest type here |
|---|---|---|
| Spear phishing | Phishing written for one person or team, after some homework | Phishing |
| Whaling | Spear phishing that goes after a senior executive | Phishing |
| Business email compromise (BEC) | Fake emails from a boss or supplier that send payments the wrong way | Phishing |
| Baiting | Leaving an infected device, or a tempting download, where the victim will find it | Physical social engineering |
| Tailgating or piggybacking | Slipping through a secure door behind an employee | Physical social engineering |
| Quid pro quo | Offering help, usually fake IT support, in return for access | Pretexting |
| Honeytrap | A fake romance or friendship used to get money or access | Pretexting |
| Watering hole | Infecting a website the target group visits a lot | ClickFix and SEO poisoning |
| MFA fatigue | Bombarding someone with push prompts until they approve one | Help-desk and MFA manipulation |
68 of the 172 cases happened in 2024 or later. Here are the newest ones.
Censys traced a USPS 'package awaiting action' smishing kit to 682 lookalike sites that streamed card numbers as victims typed. How it worked and how to tell.
A caller posing as Standard Bank got an 18-year-old to move her R438,900 education fund to a 'safe' account in 20 minutes. What happened and what's disputed.
Sysdig says an AI agent ran the JADEPUFFER ransomware attack end to end: unpatched Langflow, stolen secrets, 1,342 configs destroyed. What happened.
A Singapore businessman sent at least S$4.9M after a fake Zoom "government meeting" built from real footage of PM Lawrence Wong and other officials.
Dubai Police raided nine pig butchering centres and arrested 275, and San Diego charged six alleged scam-compound managers and recruiters. What is known.
A Pune CFO sent Rs 56 lakh after a Microsoft Teams message from a fake CEO profile. A second Rs 1.5 crore request made her call the real CEO. What happened.
Most attacks go through five stages. Here’s how they played out in the 2024 Arup case, and what would break each one.
The attackers gathered public video and audio of the company’s UK-based CFO and other staff. Hong Kong Police said it came from YouTube.
A message, supposedly from the CFO, mentioned a confidential deal and a reason to keep it quiet.
The employee joined a video call where, police said, everyone else was a pre-recorded deepfake.
Reassured, the employee made 15 transfers to five Hong Kong bank accounts over roughly a week.
Spreading HK$200 million across several accounts made it harder to freeze. Nobody noticed until the employee checked with headquarters.
Every warning sign below turned up in a real case here.
Your account is being drained, move the money now
New bank account for a known supplier
Attackers don’t need a software bug. They work on ordinary human reactions, and each one below drove a real case.
People do what the CEO or a lawyer asks. Ubiquiti’s finance staff wired $46.7M after emails that seemed to come from the CEO and an outside law firm.
A caller claiming to be Bank of America told a Chicago hairstylist that a hacker was draining her account. She wired $20,000 to the scammers.
Help desks are there to get people unstuck. MGM’s reset MFA for a caller who was only pretending to be an employee.
One fake voice is easy to doubt. A room of familiar faces is not. Arup's employee saw several "colleagues" on the call.
Send enough push prompts and eventually someone taps approve. That’s how the attacker got into Uber, through a contractor’s account.
Each control breaks one stage of an attack. The case next to it shows the kind of attack it’s built to stop.
| Control | Stage it breaks | Example case |
|---|---|---|
| Check payment requests on a channel the requester didn’t pick, using a number you already have. | Action | Arup (2024) |
| Make the help desk confirm who’s calling before it resets a password or MFA. | Contact | MGM Resorts (2023) |
| Switch to phishing-resistant MFA (FIDO2 security keys) instead of codes or push approvals. | Action | Uber (2022) |
| Have two people sign off on new payees and any change to bank details. | Cash-out | Ubiquiti (2015) |
| Train staff on phone and text lures, not just email. | Pretext | Retool (2023) |
| Keep supplier logins well away from systems that hold customer data. | Action | Target (2013) |
The median disclosed loss, across the 43 cases that state a loss figure. We give a median, not a total. The library mixes losses from single victims, campaign-wide figures, and estimates, and you can’t add those together. Each case page says what its own figure measures.
Social engineering examples are real incidents where attackers got what they wanted by fooling a person instead of breaking software: a phone call to a help desk, a fake invoice, a cloned voice, a sticker on a parking meter. This library documents 172 sourced cases, and phishing is the way in for 72 of them. The rest of this guide walks through the cases that teach the most, sorted by how the attack actually reached its victim.
Key takeaways
These social engineering attack examples aren't simply the largest. Each one shows a different lever being pulled, and each has a full, sourced case page.
| Case | Year | Attack type | What happened | Loss |
|---|---|---|---|---|
| Axie Infinity Ronin Bridge | 2022 | Spear phishing (fake job) | Engineer opened a spyware-laden "offer" after fake interviews | About $540M in crypto |
| Google and Facebook (Rimasauskas) | 2013 to 2015 | BEC, fake supplier | Invoices from a look-alike "Quanta Computer" | About $122M wired, most recovered |
| MGM Resorts | 2023 | Help-desk vishing | Caller posing as an employee got access reset | About $100M EBITDAR impact (estimate) |
| Ubiquiti Networks | 2015 | CEO fraud | Fake CEO and lawyer emails drove 14 wires | $46.7M sent, $30M not recovered |
| Arup | 2024 | Deepfake video call | Every other face on the call was fake | About US$25.6M |
| Target | 2013 | Supplier phishing | HVAC contractor's stolen login led to card theft | About 40M cards exposed |
| RSA SecurID | 2011 | Spear phishing | "2011 Recruitment Plan" spreadsheet installed a trojan | $66.3M charge (parent EMC) |
| 2020 | Vishing plus fake VPN page | Staff phoned by fake IT, 130 accounts targeted | Over $118,000 taken from the public | |
| UK energy firm | 2019 | AI voice clone | "German parent CEO" demanded an urgent payment | €220,000 (about $243,000) |
| Stuxnet at Natanz | 2007 to 2010 | USB, insiders | People carried the worm across an air gap | About 1,000 centrifuges wrecked |
The largest disclosed loss in the library started with a recruiter on LinkedIn. A senior engineer at Sky Mavis, which ran the Ronin bridge, sat through several rounds of interviews for a job at a company that didn't exist. Then came a generous offer, delivered as a document. Opening it installed spyware. North Korea's Lazarus Group went on to take 173,600 ETH and 25.5 million USDC, roughly $540 million at the time, and nobody noticed for six days. No smart-contract bug was needed. The attackers needed five of nine validator signatures and got four from Sky Mavis alone. Read the Ronin Bridge case.
Evaldas Rimasauskas didn't hack anyone. He registered a company in Latvia with the same name as Quanta Computer, a real Taiwanese supplier, opened bank accounts in that name and sent invoices that looked like every other invoice. Finance teams at two of the most technical companies on earth paid about $122 million into accounts in Latvia and Cyprus. Google got its roughly $23.3 million back. Facebook recovered most of its $98.9 million. The lesson is dull and important: a payment process that trusts a name on an invoice can be beaten by anyone who copies the name. Read the Rimasauskas case.
In September 2023, attackers reportedly rang MGM's IT help desk, claimed to be an employee and got that employee's access reset. Containing the intrusion disrupted MGM properties for about ten days. MGM estimated a hit of around $100 million to third-quarter Adjusted Property EBITDAR, which is a profit measure rather than cash stolen, plus under $10 million in one-time costs. It didn't pay a ransom. The famous details, a LinkedIn lookup and a ten-minute call, come from the attackers themselves. MGM's own filings never say how they got in. Read the MGM case.
Over two weeks, emails that seemed to come from Ubiquiti's CEO and from a real London lawyer at Latham & Watkins talked the company's own finance chief into sending 14 wires out of its Hong Kong subsidiary. The cover story was a secret acquisition, and secrecy is exactly what stopped anyone checking. The "lawyer" wrote from a generic @consultant.com address. That was the tell. Ubiquiti lost $46.7 million, clawed back $16.7 million and never saw the remaining $30 million. It found out only when the FBI emailed its CEO. Read the Ubiquiti case.
A finance employee in Arup's Hong Kong office suspected the first message was phishing. Fair enough. Then came a video conference with what looked like the UK chief financial officer and several colleagues, and the doubt went away. Hong Kong police say everyone on the call except the employee was a deepfake built from public video, and the clips were pre-recorded rather than live conversation. The employee made 15 transfers totaling HK$200 million, about US$25.6 million, to five local bank accounts. No recovery has been publicly reported. Read the Arup case.
Target's breach began at Fazio Mechanical Services, a small refrigeration and HVAC contractor in Pennsylvania. A malware-laden email at the contractor stole its login to Target's vendor systems. With that legitimate credential, the attackers looked like a trusted partner, moved into Target's network and planted card-stealing malware on checkout terminals. Between 27 November and 15 December 2013, about 40 million payment cards were exposed, along with personal details for up to 70 million customers. The weak point was a supplier with far less security than the retailer it served. Read the Target case.
A security company, beaten by a spreadsheet. In March 2011, RSA staff got an email titled "2011 Recruitment Plan". At least one copy landed in a junk folder, and an employee pulled it out and opened the Excel attachment anyway. An unpatched Flash flaw installed a remote-access trojan, and the attackers worked their way to systems holding SecurID data. RSA later said the stolen information was used in an attempted attack on Lockheed Martin. Its parent, EMC, took a $66.3 million charge. The targets weren't executives. They were ordinary staff. Read the RSA case.
In July 2020, callers posing as Twitter's IT help desk told employees there was a VPN problem and steered them to a fake login page. Staff typed in passwords and approval codes, and the attackers climbed into internal admin tools. They went after 130 high-profile accounts, including Barack Obama, Joe Biden, Elon Musk, and Apple, and posted a "double your bitcoin" scam that took over $118,000 from the public. No software flaw was used. App-based MFA didn't help either, because staff approved the prompts themselves. Read the Twitter case.
A Friday afternoon in March 2019. The head of a UK energy company takes a call from his boss, the chief executive of the German parent. The accent is right, the rhythm is right. €220,000, about $243,000, has to reach a Hungarian supplier within the hour. He pays. Then a second call asks for another payment, claiming the first was already reimbursed. It wasn't, and the follow-up call came from an Austrian number. This is one of the earliest reported cases of AI voice cloning used for CEO fraud. Read the voice clone case.
Iran's Natanz enrichment plant was air-gapped. The gap held. People simply walked across it. Stuxnet reached the plant on USB drives, carried first by a recruited insider posing as a mechanic and later by contractors who had no idea what they were holding. It went on to wreck about 1,000 IR-1 centrifuges. The attack is widely attributed to a joint US and Israeli program called "Olympic Games", though neither government has confirmed it, so treat the attribution as alleged. Read the Stuxnet case.
Email and fake login pages remain the most common way in. What changes is the bait. Sometimes it's an account warning, sometimes a document someone is expecting.
More cases on the phishing hub.
Business email compromise rarely involves malware. It involves a plausible email, a payment request and someone who doesn't want to slow down a boss. The numbers get large fast.
See also business email compromise and spear phishing, plus the phishing hub.
A phone call carries tone, urgency and a live person who can answer objections. That's why help desks and bank customers keep getting hit.
The MGM and Twitter cases above belong here too. More on the vishing hub.
Texts are short, so the scam has to be short. Unpaid toll. Missed parcel. IT needs you to log in.
More on the smishing hub.
A QR code hides the link and moves the click to a phone, which usually has weaker protection than a work laptop.
More on the quishing hub.
Synthetic media makes an old trick, pretending to be the boss, much harder to dismiss. Several of the best cases are ones that failed, and they show why.
Arup and the UK energy firm, above, are the two costliest examples. Browse the deepfake hub or the voice cloning page.
Pretexting is the story itself. A believable role, a reason to act, and enough detail to make questions feel rude.
More on the pretexting hub.
Doors, badges and free gifts. Physical social engineering is old, cheap and still works.
Stuxnet, above, is the most consequential USB case on record. More on the physical social engineering hub.
Here the victim runs the attacker's code themselves, believing it's a fix or a security check. Or they find the bait through a search engine.
More on the ClickFix hub.
The newest cases flip the script. Sometimes the AI is the attacker's assistant. Sometimes the AI is the one being conned.
More on the agentic AI hub.
Read enough real social engineering examples and the same shapes repeat. The statistics page counts them across the whole library.
In 2024, a finance employee at Arup in Hong Kong joined a video call with what looked like the company’s CFO and colleagues. According to Hong Kong Police, every other person on the call was a deepfake. The employee made 15 transfers totaling HK$200 million, about US$25.6 million, before checking with headquarters.
The four most common in this library are phishing (72 cases), vishing (27), physical social engineering (22), and pretexting (22). Phishing uses email, vishing uses phone calls, physical attacks happen in person, and pretexting builds a false story before any request.
Phishing. It appears in 72 of the 172 cases here (42%), more than any other type. Email is still the main channel, but 39 cases in this library involved a phone call or a text message.
Scenarios that recur in this library include a fake IT support call asking for a password or MFA reset, a fake executive or supplier asking for an urgent payment, a text message with a link to a fake login page, and a fake bank calling about fraud on your account.
Hacking exploits a flaw in software. Social engineering exploits a person, so there is no patch for it. That is why a convincing call to the help desk can get an attacker an MFA reset at an organization whose systems are fully up to date.
Every case cites its sources, with primary documents such as court filings, regulator notices, and company disclosures listed first. Each case is labeled as a documented incident, a reported or alleged incident, or research. The methodology page explains the rules.
172 cases, each showing who was hit, how the attacker got in, what it cost, and the sources. Search by company, attack type, or year.