
Microsoft calculated that enabling MFA reduces the risk of account takeover by 99.22%. Proofpoint calculated something else: in 59% of companies where criminals managed to take over accounts, the victims had MFA enabled. And Sophos adds that in 97% of ransomware attacks that started with stolen credentials, MFA was deployed in some form. All of these numbers are true. And that is exactly what this post is about. 💥
When I ask during pentests, “Do you have 2FA?”, these days I almost always hear “Yes, of course.” Far less often does anyone ask back: what kind? And that makes all the difference. An SMS code, an app that generates codes, an “Approve sign-in?” notification and a passkey are four completely different levels of protection. Some of them a criminal bypasses with a ready-made phishing kit rented on a subscription, like Netflix – $120 for 10 days (more on that below, under Tycoon 2FA).
So I gathered hard data: effectiveness studies (Microsoft, Google), adoption statistics, real breaches despite 2FA (Uber, MGM, Marks & Spencer, Retool), the scale of SIM swaps in the US and in Poland, Polish data from CERT Polska (Poland’s national CERT) and the NBP (the National Bank of Poland), and the state of passkeys in November 2026. Every number has a source and a date (list at the end). Finally, a table showing which method falls to which attack, and 7 things to do right away.
Table of contents
- First, the good news: MFA really works
- How many people and companies actually use MFA?
- SMS codes: SIM swaps, SS7 and the companies your codes pass through
- Authenticator app codes: phishing that relays your 2FA live
- Push notifications: whoever gets tired first loses
- When MFA doesn’t matter: session theft and infostealers
- The weakest link: a phone call to the help desk
- Passkeys and hardware keys: are they really impossible to bypass?
- Poland: where the money actually disappears
- Ranking: which method falls to which attack
- 7 things to do right away
- FAQ
First, the good news: MFA really works
For non-technical readers: MFA (multi-factor authentication), commonly called 2FA, is the principle that “a password alone is not enough.” On top of the password (something you know) comes something you have (a phone, a key) or something you are (a fingerprint).
Before I start pointing out weaknesses, let’s give the numbers their due:
- Microsoft, 2023: a study of Azure AD (now Entra ID) accounts found that MFA reduces the risk of account takeover by 99.22%, and for accounts whose passwords had already leaked, by 98.56%. The authenticator app performed better than SMS, but both were incomparably better than no MFA at all.
- Microsoft Digital Defense Report 2026 (October 2026): password spraying (trying common passwords against thousands of accounts at once) still makes up more than 99% of observed intrusion attempts against the cloud. Practically any MFA is enough to stop these attacks. The good news: password attacks dropped by 26% year over year. The bad news: the share of MFA-bypassing phishing and token theft more than doubled in the first half of 2026 – although it is still relatively small.
- Google, 2019: the most frequently cited study broken down by method. Chart below.

Look at the last column in each group. An SMS code blocks all bots, but one in four targeted attacks still gets through. The hardware key in this study: zero victims.
Lesson #1: Any MFA is a huge leap in security, because the vast majority of attacks are mass attacks that bounce off the second factor. But “any MFA” protects you from the crowd, not from someone who came for you. And the number of attackers who come for someone specific is growing – more on that below.
How many people and companies actually use MFA?
The weakest MFA is the one you don’t have. And here the numbers are sobering:
- Microsoft Entra ID (February 2024): 38% of monthly active users used MFA (in September 2022: 26.6%). We’re talking about business accounts in Microsoft’s cloud.
- Okta (December 2025 report, January 2025 data): 70% of employees at companies using Okta have MFA. So almost one in three still signs in with a password alone. Technology does best (87%), transportation and warehousing worst (42%).
- X (Twitter), second half of 2021: 2.6% of active users had 2FA enabled, and 74.4% of them used SMS. That is the last such data the platform has published.
- FIDO Alliance (May 2026): 57% of companies still rely on phishable methods for employees’ everyday sign-in. Microsoft quoted the same figure in November 2026 at its Ignite conference.

The trend is good: phishing-resistant methods grew by 63% in a year (from 8.6% to 14% of users), and SMS is slowly declining. But 93% of people still use a password.
Lesson #2: Before you start wondering whether your MFA is “good enough,” check whether it’s everywhere. On engagements, I most often get in not through broken 2FA, but through that one forgotten account: a service account, a test account, an old VPN panel or an admin account from before the migration that “was supposed to be disabled.”
SMS codes: SIM swaps, SS7 and the companies your codes pass through
SMS is the most popular second factor in the world and, at the same time, the weakest. The US agency CISA bluntly calls it a “last resort,” and the new version of the NIST guidelines (SP 800-63B-4, July 2025) keeps SMS and voice codes in “restricted” status: anyone who uses them must offer users a more secure alternative, warn them about the risk and have a migration plan.
Microsoft goes further: the SMS codes and voice calls it sends in Entra ID (that is, for signing in to Microsoft 365 at work) are set to disappear from February 1, 2027 for most users and from July 1, 2027 for administrators. They will remain only for companies that connect their own SMS provider. The same will apply to password resets.
Why is everyone running away from SMS? Because an SMS can be intercepted in at least three ways before it even reaches the victim.
1. SIM swap: the criminal gets your number
The criminal convinces your mobile carrier (or bribes one of its employees) to move your number to their SIM card. From that moment on, all your codes go to them. All you see is “No service.”
- FBI IC3, 2025: 971 reported SIM swaps, $17.4 million in losses (2024: 982 reports, $26 million). And those are only the cases reported in the US.
- Scattered Spider, November 2026: Tyler Buchanan, the last of five defendants charged in the US, was sentenced to 6 years in prison. The group combined SMS phishing with SIM swaps to bypass 2FA; his case involved $11.6 million.
- Poland, 2018-2021: a group broken up by police in Poznań obtained duplicate SIM cards using stolen personal data and forged documents. More than 50 victims, nearly PLN 20 million siphoned off and an attempt to siphon off another PLN 26 million. In 2024, 12 people were indicted.
- Poland, June 2026: the CBZC (Poland’s Central Cybercrime Bureau), together with the FBI, arrested 4 people who, according to investigators, broke into the systems of companies working with mobile carriers, hijacked phone numbers and reset passwords on cryptocurrency exchanges. Losses ran into the millions.
Regulators are starting to react: in the US, FCC rules took effect on November 17, 2026 requiring carriers, before swapping a SIM card or porting a number, to verify the customer without relying on biographical data such as a social security number or date of birth, to notify them of the change and to offer a free account lock.
2. SS7: a hole in the very heart of the mobile network
SS7 is a protocol from the 1970s that carriers all over the world use to exchange information about traffic. Whoever gains access to it can redirect someone else’s text messages. This isn’t theory: in 2017 in Germany, criminals used access at a foreign carrier to redirect, at night, the bank authorization codes of O2 customers, and in 2019 the British Metro Bank confirmed a similar attack.
3. Middlemen you’ve never heard of
When Google, WhatsApp or your bank sends you a code, the SMS usually doesn’t come straight from them. It passes through intermediary companies. In June 2025, Lighthouse Reports and Bloomberg described the Swiss company Fink Telecom Services, which handled sign-in messages for more than 1,000 companies. The leak obtained by the journalists contained millions of messages, including sign-in codes for Google, Meta, Amazon, WhatsApp, Signal and banks.
Then there are ordinary OTP bots on Telegram: an automated system calls the victim pretending to be “the bank,” asks them to “confirm their identity with a code” and, at that very moment, signs in to their account.
Lesson #3: SMS protects you from bots and mass phishing, but not from anyone who has your name and a bit of patience. If a service gives you a choice, choose anything else. If it doesn’t, ask your carrier for an extra customer service password and ask how it protects the issuing of a duplicate SIM card.
Authenticator app codes: phishing that relays your 2FA live
An app like Google Authenticator or Microsoft Authenticator generates the code (TOTP) locally on your phone. SIM swaps and SS7 stop working. Great. Except there is one attack against which an app code protects you just as poorly as SMS: adversary-in-the-middle (AiTM) phishing.
How does it work? The fake page isn’t a copy of the sign-in page, it’s a relay. You type your password – the page passes it live to the real Microsoft. Microsoft asks for a code – the page asks you. You type the code – it passes it on. The sign-in succeeds, and the criminal gets the session cookie, a digital pass that tells the server “this person has already authenticated.” From then on, they don’t need your password or your code at all. 😈
The most devious detail: after “signing in,” you don’t see any error. The relay shows you your real inbox, real emails, real photos – because it really is signed in. Everything looks normal, so nobody changes their password or calls IT.

The scale of the problem:
- Microsoft Digital Defense Report 2026: AiTM already accounts for 44.6% of identified phishing techniques. Two years earlier, Microsoft reported a 146% year-over-year increase in such attacks.
- Tycoon 2FA: a ready-made phishing kit sold as a subscription service, starting at $120 for 10 days or $350 a month. In mid-2025 it was responsible for about 62% of all phishing blocked by Microsoft. On March 4, 2026, Microsoft, together with Europol, seized 330 of the kit’s domains; according to Microsoft, Tycoon had about 96,000 victims, including more than 55,000 Microsoft customers. The police operations also covered Poland, among other countries. The service came back after a few days, but by June 2026 its activity had dropped by 95% from its peak – because the operators moved on to a newer trick.
- That newer trick is device code phishing. The victim receives a code and a link to the real Microsoft page, enters the code there and approves the sign-in themselves – except they are signing in the criminal’s device. No fake sign-in page is needed at all. In September 2026, Microsoft took down the EvilTokens service, which had hijacked more than 12,000 mailboxes this way; two people were arrested in London.
- Proofpoint (data from November 2024 to November 2025): criminals tried to take over accounts in 99% of the organizations studied and succeeded in 67%, and in 59% of companies with compromised accounts, the victims had MFA enabled.
A fun fact from my own backyard: the most popular tool for these attacks in penetration testing, Evilginx, was written in 2017 by a Pole, Kuba Gretzky. As a tool for education and testing. Today it’s used both by pentesters (including me, on engagements) and by criminal groups like Scattered Spider. In an interview with The Record, the author openly admitted that he is aware he indirectly helps the “bad guys” too, which is why he now gives the full version only to vetted companies.
Cat and mouse: how services detect relays
- A script on the sign-in page checks the address in the browser bar and sends it along with the password. If someone signs in to Facebook from a domain other than facebook.com, the server knows it’s an attack.
- Public certificate logs. Every site with a padlock needs a certificate, and every certificate issued ends up in a public log. Security companies track new domains and catch the ones that look like a bank sign-in page.
- The criminals’ answer: a secret suffix in the link. The link in the SMS or email ends with a random string of characters. Anyone who visits without it – a security scanner, for example – sees an error or gets redirected to a harmless page. The victim sees the fake sign-in page.
Every protection gets analyzed and bypassed. If you want to see how such a relay works from the inside, I recommend Kacper Szurek’s video “How hackers steal 2FA-protected accounts” (in Polish) – brand new.
One more story, this time about an app: in August 2023, Retool lost control of the accounts of 27 cloud customers. The attacker impersonated the IT department over a voice call and tricked an employee into handing over a code. And because Google Authenticator synced codes with the Google account, compromising that one account gave access to all the codes at once. Retool gave its report a telling title: “When MFA isn’t actually MFA.”
Lesson #4: Every code you type in or read out – from an SMS, an app or an email – can be relayed in real time. An authenticator app protects you from SIM swaps, but not from a fake page. The ultimate solution isn’t better relay detection, but a method that can’t be relayed, because it checks by itself which site it’s on. More on that in a moment.
Push notifications: whoever gets tired first loses
A notification asking “Are you trying to sign in? Yes / No” seems like the most convenient option. And it is. Including for a criminal who knows your password: all they have to do is keep sending notifications until someone taps “Yes” just to make the phone stop buzzing. That’s MFA fatigue, an attack that wears you down.
- Microsoft, 2022: about 1% of users approve a plain notification on the very first try, without thinking. With thousands of employees, there is always someone.
- Cisco, May 2022: attackers got an employee’s password from a personal Google account, then combined calls “from IT support” with a barrage of notifications. The employee approved one, and the attackers got into the VPN.
- Uber, September 2022: a contractor’s password bought online, then a barrage of notifications and a WhatsApp message “from Uber IT.” The contractor eventually approved. The attacker got into, among other things, Slack, G-Suite and the HackerOne panel with vulnerability reports.
The industry’s answer is number matching: a number appears on the sign-in screen, and you have to enter it in the app. You can’t approve on reflex. Microsoft enforced this for all customers on May 8, 2023. Problem solved? Partly. Number matching doesn’t protect against AiTM, because the fake page simply shows you the real number. CISA still classifies push with number matching as a phishable method.
Lesson #5: If your company uses push notifications without number matching – turn it on today, it’s a single setting. And tell employees clearly: a notification you weren’t expecting means someone knows your password. You tap “No,” change your password and call IT. Not the other way around.
When MFA doesn’t matter: session theft and infostealers
Everything I’ve described above are attacks on the moment of sign-in. But what if the criminal doesn’t need to sign in at all?
Infostealers are malicious programs that, once they infect a computer (most often through “free” software, cracks or fake updates), pull saved passwords and active session cookies out of the browser. With such a cookie, the criminal is “already signed in.” Your MFA did its job in the morning, when you signed in. For them, it no longer matters.
- Recorded Future (March 2026): in 2025, 892 million stolen credentials were exposed, 276 million (31%) of them with active session cookies.
- Lumma Stealer: before Microsoft and law enforcement seized about 2,300 of its domains in May 2025, it had infected more than 394,000 Windows computers in just two months.
- Snowflake, 2024: attackers got into the environments of up to 165 of the platform’s customers using passwords stolen by infostealers, sometimes several years earlier. The common denominator: accounts without MFA.
- Okta, October 2023: an attacker got into Okta’s customer support system and pulled out diagnostic files (HAR) that customers had uploaded to support. They contained session tokens. The attacker used them to target 5 customers, including 1Password, BeyondTrust and Cloudflare.
Lesson #6: MFA protects the door, but not against someone who stole the key to a room that’s already open. That’s why, alongside MFA, you need: shorter sessions for critical systems, binding sessions to the device (e.g. Conditional Access policies in Entra ID), monitoring of what happens after sign-in and, above all, a ban on installing “just anything” on work computers. Most infections start with a single pirated game or a “free” PDF converter.
The weakest link: a phone call to the help desk
The most expensive breaches of recent years required no technical skill at all. They required one phone call to the IT department: “I lost my phone, please reset my MFA.”
- MGM Resorts, September 2023: the Scattered Spider group impersonated an employee in a call with the help desk. The impact, according to MGM’s filing with the US SEC: a hit of about $100 million to the quarter’s results.
- Clorox, 2023: according to Clorox’s lawsuit against Cognizant (July 2025, a $380 million claim), the attacker called the help desk and, without any verification, got a password reset, an MFA reset and a change of the phone number used for SMS codes. Several times.
- Marks & Spencer, April 2025: according to the company’s chairman, speaking before the UK Parliament, the attackers got in through “sophisticated impersonation” and a password reset. Expected hit to profit: about £300 million. Co-op, attacked at the same time, lost £80 million in profit for the half-year, and the data of all 6.5 million of its members leaked.
Members of Scattered Spider are largely teenagers and people in their twenties. Sentences were handed down in 2025 and 2026, but the playbook still works – in September 2026, Microsoft described further groups that call and message employees with the line “update your passkey” and then register their own MFA methods on the victims’ accounts.
Lesson #7: Your MFA is only as strong as its recovery procedure. If a reset only takes a first name, last name and department, you have MFA on paper. Verification at reset should be at least as strong as the sign-in itself: a callback to a known number, confirmation from a manager, a video call with an ID document. It’s one of the things I always check in social engineering tests – and, reading between the lines, it doesn’t always come out well.
Passkeys and hardware keys: are they really impossible to bypass?
A passkey (Polish services often call it a “klucz dostępu,” an access key) is the successor to the password, based on the FIDO2/WebAuthn standard. Instead of typing a code, you confirm the sign-in with your fingerprint, face or PIN on your own device. The key difference: a passkey is bound to a specific domain. On a fake site like mlcrosoft-login.com it simply won’t work, because your phone has no key for it. There is nothing to type in, so there is nothing to phish. ✔️
A passkey can be synced (iCloud, Google Password Manager, password managers) or device-bound, e.g. a hardware key such as a YubiKey (about $58-85 from the manufacturer).
Real-world results
- Google, 2018: since early 2017, when the company mandated hardware keys for more than 85,000 employees, not a single successful phishing attack on work accounts has been recorded.
- Cloudflare, July 2022: more than 76 employees received an SMS with a link to a fake sign-in page, and 3 entered their credentials. The attack still failed, because every employee signs in with a FIDO2 key. The same campaign at other companies, including Twilio, ended in breaches.
- FIDO Alliance (October 2026): about 7 billion passkeys are already in use worldwide. At the large consumer services in the FIDO index, passkeys already account for 36% of successful sign-ins (26% a year earlier), and signing in with a passkey succeeds in 93% of attempts, compared with 78% for a password with traditional MFA.
- Microsoft: in May 2025 it reported a 98% sign-in success rate for passkeys versus 32% for passwords, and in November 2026 it announced passkeys as the default method for new Entra ID accounts.
Where passkeys have loopholes
I wouldn’t be a pentester if I didn’t go looking for a hole. Here’s what is actually known:
- Downgrade attack: in 2025, Proofpoint showed that a fake page can pretend to be a browser that “doesn’t support” passkeys and force the victim to sign in with a weaker fallback method. In the video mentioned above, Kacper Szurek describes an even smarter variant: a virtual browser runs on the criminal’s server and, when it hits a key-protected account, clicks “Cancel” and picks a phone notification instead. The victim – convinced they are signing in on their own – approves it and enters the correct number. The key stays safely in their pocket, and the account is taken over anyway. All of these variants only work when the account has a weaker fallback method.
- Registering someone else’s passkey: since a passkey can’t be stolen or phished, criminals talk the victim into… adding their key. In 2026, Okta and Unit 42 described campaigns in which “the IT department” calls asking you to “register a new passkey,” and Google Threat Intelligence Group described the UNC6671 group, which attacked more than 200 companies using the line “enable FIDO2 passkeys.”
- Infected computer: in August 2026, Unit 42 (Palo Alto Networks) researchers described how malware on Windows can extract a secret that makes it possible to recreate passkeys synced through Google Password Manager. At Black Hat 2026, researchers also showed a Windows bug (CVE-2026-34348) that allowed intercepting and replaying a passkey sign-in to Entra ID – patched in July 2026. The moral: a compromised computer is a compromised computer, whatever the sign-in method.
- Browser extensions: at DEF CON 33 (August 2025), Marek Tóth demonstrated clickjacking of password manager extensions – in 8 of the 11 tested, passkey sign-in could be abused, and in 9 of 11, TOTP codes could be stolen. Some vendors patched it quickly, others classified it as an informational risk.
- Physical attack on the key: EUCLEAK (2024) made it possible to clone older YubiKey 5 keys (firmware below 5.7), but it required physical access to the key, taking it apart and expensive lab equipment. Not a realistic scenario for hijacking your online shopping account.
Notice the pattern: none of these attacks “breaks” passkey cryptography. They get around it through a fallback method, registering a new key, a compromised computer or the account recovery process. Exactly the same weak spots as with every other method.
Lesson #8: Passkeys and hardware keys are currently the only methods that actually stop AiTM phishing. But only if you disable weaker fallback methods and control who can add a new key. A passkey next to an SMS code is a high-security lock on a door with a glassless window right beside it. For administrators and the board: a hardware key (at its highest level, AAL3, NIST explicitly requires non-synced keys), and on a Google account, the Advanced Protection Program, which requires keys and switches off weaker methods.
Poland: where the money actually disappears
In Poland, the problem looks a little different from Microsoft’s corporate reports. Because here, most of the money is lost not by bypassing MFA, but because the victim approves the transaction themselves.
- CERT Polska, 2025 report: 260,783 incidents, 97% of them computer fraud. About 78,000 phishing incidents, most often targeting users of OLX and Allegro (Poland’s biggest online marketplaces). 244,341 domains were added to the warning list (166% more than a year earlier), and mobile carriers blocked about 1.88 million scam text messages.
- NBP, Q1 2026: fraudulent transfers worth more than PLN 245 million – a new record since the NBP began measuring, over a third higher than the previous one (nearly PLN 177 million in Q3 2025).
- NBP, Q2 2026: 24,400 fraudulent transfers worth PLN 142.9 million, of which 80.1% by value was “payer manipulation”: the victim orders the transfer themselves or hands over their authorization data themselves.
- The “bank employee” scam: in December 2025, the CBZC described an organized call center that impersonated bank hotlines. Documented losses in just one case: at least PLN 6 million, and in total, according to estimates cited by the Polish business portal wnp.pl, as much as several hundred million.
- Fresh campaigns (CERT Polska, September and October 2026): Telegram account takeovers via an SMS about “suspicious activity,” leading to a page that collects the phone number and then the verification code – the campaign was still running in October. In the same month, fake iPKO pages (the online banking of PKO BP, Poland’s largest bank) were circulating with a lure about “new transfer authorization rules.” A classic: 2FA works, it’s just the victim who enters the code.

The state sees it too. On November 17, 2026, Poland’s Government Plenipotentiary for Cybersecurity – together with UKNF (the Office of the Polish Financial Supervision Authority) and CSIRT KNF – issued a recommendation for the banking sector (No. 2026-67a-10). It states that for privileged, administrative and remote access accounts, organizations should use multi-factor authentication that is “phishing-resistant, where the technology in use allows it,” and that identity monitoring should also cover what happens after sign-in, because hijacking a token or an active session makes it possible to act “without having to break the authentication again” (quotes translated from Polish). That is exactly what this post is about.
What does security look like in Polish services in November 2026?
- Sign-in with a FIDO hardware key is offered by Polish banks including ING (the first in Poland, since summer 2023), PKO BP (in iPKO since September 2024) and Pekao (since November 2024).
- Passkeys: BNP Paribas launched them in October 2026 in its GOonline Biznes corporate banking. mBank added a “digital access key” to its terms and conditions, effective December 22, 2026, but hasn’t announced a launch date yet. Pekao plans passkeys for card payments from Q4 2026. Allegro has had passkeys since April 2026, and Revolut supports them too.
- Profil Zaufany (the Polish government’s e-ID login): two-factor sign-in mandatory since July 2023 – an SMS code or confirmation in the mObywatel app (the government’s digital ID app). I didn’t find any support for FIDO keys.
- SMS codes as a fallback method are still available at practically every bank.
Lesson #9: In Polish reality, the second factor that fails most often is the human. No method will help if the customer approves the transfer in the app themselves because “a bank employee” told them to. A passkey has one big advantage here: it simply can’t be used on a fake bank page. So turn on a key at your bank if it offers one. And there is only one answer to a call “from the bank”: hang up and call the number from your bank’s official website yourself.
Ranking: which method falls to which attack
Compiled from CISA and NIST guidelines and the attacks described above. ⚠️ means “the attack works,” ✔️ means “the method protects.”
| Method | AiTM phishing (live fake page) | SIM swap / SS7 | Fatigue attack (push) | Session theft (infostealer) | Help desk reset |
|---|---|---|---|---|---|
| SMS code | ⚠️ | ⚠️ | – | ⚠️ | ⚠️ |
| Email code | ⚠️ | ✔️ | – | ⚠️ | ⚠️ |
| App code (TOTP) | ⚠️ | ✔️ | – | ⚠️ | ⚠️ |
| Push without number matching | ⚠️ | ✔️ | ⚠️ | ⚠️ | ⚠️ |
| Push with number matching | ⚠️ | ✔️ | ✔️ | ⚠️ | ⚠️ |
| Synced passkey | ✔️ (as long as there is no weaker fallback method) | ✔️ | ✔️ | ⚠️ | ⚠️ (depends on the procedure) |
| FIDO2 hardware key | ✔️ (as long as there is no weaker fallback method) | ✔️ | ✔️ | ⚠️ (the session yes, the key itself no) | ⚠️ (depends on the procedure) |
In the last two columns, there is a warning next to every method, and that is the most important takeaway from this table: session theft and help desk resets bypass every sign-in method, because they attack what happens before or after it. The same goes for device code phishing: there, the victim signs in on the real page themselves.
The ranking, from weakest to strongest:
- No MFA – an open door.
- SMS code / voice call – better than nothing, a “last resort” according to CISA.
- Email code – depends on how well the email account is protected.
- Push without number matching – convenient, but vulnerable to fatigue attacks.
- App code (TOTP) and push with number matching – a solid middle class, but vulnerable to AiTM phishing.
- Synced passkey – phishing-resistant, ideal for most people.
- FIDO2 hardware key / device-bound passkey – for administrators, the board, finance and anyone who is a target.

7 things to do right away
- Turn on MFA everywhere you can – start with your email. Your email is the key to resetting the passwords of all your other accounts. Then your bank, password manager, work accounts and social media.
- Wherever you can, switch to passkeys. Google, Microsoft, Apple, Amazon, Allegro, some banks. It takes a minute. Have a hardware key and a Google account? Turn on the Advanced Protection Program.
- Remove SMS as a fallback method if you already have a passkey or an app. Keep your backup codes written down on paper, in a safe place. And never add a passkey or enter a device code because “the IT department” asked you to over the phone.
- Secure your phone number with your carrier: set an extra customer service password and ask how the carrier protects the issuing of a duplicate SIM and number porting.
- At work: hardware keys for administrators and the board, number matching for everyone, and in Entra ID / Google Workspace a policy that requires phishing-resistant methods for privileged accounts. If nobody uses device code sign-in – block it. And plan your move away from SMS before February 2027.
- Review your help desk’s MFA reset procedure. Call in yourself, posing as an employee, and see what they ask you for. If a first name, last name and department are enough – you have a problem.
- Test it for real. A phishing exercise with an AiTM kit and a social engineering test of your help desk will show you within a few days what your MFA is really worth. If you want to know how much such a test costs, check the calculator, and if you’d rather talk – get in touch.
FAQ
Is SMS 2FA secure?
It is much better than a password alone – according to Google’s 2019 study, an SMS code blocks 100% of bot attacks and 96% of mass phishing. But it lets one in four targeted attacks through and is vulnerable to SIM swaps, attacks on the SS7 network and live phishing. CISA describes SMS as a “last resort,” NIST as a restricted method, and from February 2027 Microsoft will stop sending SMS codes itself in Entra ID.
Is an authenticator app (Google Authenticator, Microsoft Authenticator) better than SMS?
Yes, because the code is generated on your phone and can’t be intercepted through a SIM swap or SS7. It can still be phished, though, through a fake page that relays data live (AiTM phishing, e.g. Tycoon 2FA kits or Evilginx).
What is an MFA fatigue attack?
It means bombarding the victim with “Approve sign-in?” notifications until they finally tap “Yes.” That is how Uber and Cisco were breached in 2022. The defense is number matching, which Microsoft enforced for all customers in May 2023.
What is device code phishing?
The criminal sends the victim a code and a link to the real Microsoft sign-in page. The victim enters the code and goes through normal MFA, but in doing so signs in the criminal’s device. No fake page is needed, so a passkey won’t help either. Protection: block this sign-in method in your company if nobody needs it, and follow the rule “I never enter codes someone asks me for.”
Can passkeys be hacked?
So far, nobody has broken passkey cryptography in practice. Passkeys can, however, be bypassed: through a weaker fallback method on the same account, by talking the victim into registering the criminal’s key, through a compromised computer (malware, browser extensions) or through the account recovery process. That is why, after enabling a passkey, it’s worth removing SMS as a fallback method.
Which 2FA method is the most secure?
A FIDO2 hardware key or a device-bound passkey, with weaker fallback methods disabled. Since 2017, Google has not recorded a single successful phishing attack on the accounts of employees who use hardware keys.
Does MFA protect against session theft?
No. An infostealer steals a session cookie from the browser that was issued after sign-in, so the criminal doesn’t have to go through MFA at all. According to Recorded Future, 31% of credentials stolen in 2025 contained active session cookies. Protection means shorter sessions, binding sessions to the device, monitoring activity after sign-in and preventing infections in the first place.
Which Polish banks support U2F/FIDO2 keys and passkeys?
As of November 2026: sign-in with a hardware key is offered by banks including ING, PKO BP and Pekao. BNP Paribas has introduced passkeys in its corporate banking. mBank added a “digital access key” to its terms and conditions effective December 22, 2026, but will announce the launch date later. SMS codes as a fallback method are still available at practically every bank.
Summary
“We have 2FA” is no longer enough to sleep soundly. The numbers show three levels. Any MFA stops mass attacks, which make up the vast majority of attacks – according to Microsoft, password guessing alone accounts for more than 99% of attempts to break into the cloud. Codes and notifications – via SMS, app or push – fall to live phishing that anyone can rent for $120. Passkeys and hardware keys stop that too, as long as you don’t leave a back door next to them in the form of SMS and don’t let anyone add their own key. And on top of all that are the attacks that bypass every method: session theft, a device code entered because “IT” asked for it, and one convincing phone call to the help desk. That’s where you need a lock too. 🔓
If you want to find out whether your company’s MFA will withstand an AiTM kit, a call “from the IT department” or a well-crafted SMS – you know where to find me.
Sources of the data used in this post
- Microsoft Research – Meyer et al., “How effective is multifactor authentication at deterring cyberattacks?” (arXiv:2305.00945, 05.2023)
- Microsoft Digital Defense Report 2026 (10.2026) and 2024 (10.2024, CISO summary)
- Sophos – State of Ransomware 2026 (07.2026)
- Google Security Blog – “New research: How effective is basic account hygiene at preventing hijacking” (05.2019)
- Microsoft Security (X) / Practical 365 – MFA adoption in Entra ID (02.2024)
- Okta – Secure Sign-in Trends Report 2025 (12.2025)
- X / Twitter – Account Security transparency report (second half of 2021); The Hacker News (02.2023)
- FIDO Alliance – “The State of Passkeys 2026” (05.2026), “7 billion passkeys and counting” (10.2026) and Passkey Index 2026 (Biometric Update, 10.2026)
- Microsoft Entra Blog – Ignite 2026 (11.2026); Microsoft 365 Message Center MC1426371 and MC1536049 – retirement of SMS and voice calls in Entra ID
- NIST SP 800-63B-4 (07.2025); CISA – “Implementing Phishing-Resistant MFA” (10.2022)
- FBI IC3 – Internet Crime Report 2024 and 2025
- SecurityWeek and Malwarebytes – Tyler Buchanan’s sentencing (11.2026); CommLaw Group – FCC rules on SIM swaps (11.2026)
- Telepolis (07.2023), CRN (07.2024) – the Poznań SIM swap group; BleepingComputer / Security Bez Tabu (06.2026) – CBZC arrests
- Help Net Security – SS7 and O2 (05.2017); Vice / Motherboard – Metro Bank (01.2019)
- Lighthouse Reports – “Two Factor Insecurity” (06.2025)
- Krebs on Security – OTP bots (09.2021)
- Microsoft On the Issues and Microsoft Security Blog – Tycoon 2FA (03.2026); BleepingComputer – the return of Tycoon 2FA (03.2026); Okta – dispersal of Tycoon 2FA operators (05.2026); The Hacker News – EvilTokens (09.2026)
- Proofpoint – “Account compromise in the agentic workspace” (02.2026) and “FIDO authentication downgrade” (08.2025)
- The Record – interview with Kuba Gretzky (10.2025)
- Kacper Szurek – “Jak hakerzy kradną konta zabezpieczone 2FA” (YouTube, in Polish, 11.2026)
- Retool – “When MFA isn’t actually MFA” (09.2023)
- Redmondmag – Microsoft on MFA fatigue attacks (09.2022); BleepingComputer – number matching enforcement (05.2023)
- Uber Newsroom – security update (09.2022); The Hacker News – Cisco (08.2022)
- Recorded Future – Identity Trend Report (03.2026)
- Microsoft On the Issues – Lumma Stealer (05.2025); The Hacker News – Snowflake (06.2024); BleepingComputer – Okta (11.2023)
- MGM Resorts – Form 8-K filed with the SEC (10.2023); The Register – Clorox vs Cognizant (07.2025) and Co-op (09.2025); BleepingComputer – M&S (07.2025)
- Krebs on Security, SecurityWeek, The Hacker News – Scattered Spider sentences (2025-2026); The Hacker News – “update your passkey” campaigns (09.2026)
- Security Boulevard – passkey registration by attackers (Okta, Unit 42; 11.2026); Security Affairs – UNC6671 (08.2026)
- Krebs on Security – hardware keys at Google (07.2018); Cloudflare – “The mechanics of a sophisticated phishing scam” (08.2022)
- Microsoft Security Blog – “Pushing passkeys forward” (05.2025)
- The Hacker News – Unit 42 and SpecterOps on passkey attacks (08.2026); Socket – password manager clickjacking, Marek Tóth (08.2025); NinjaLab – EUCLEAK (09.2024)
- CERT Polska – 2025 report (NASK, CyberDefence24, 04.2026) and monthly summaries for 09.2026 and 10.2026
- NBP – information on fraudulent transactions for Q1 and Q2 2026 (Cashless, 07.2026 and 11.2026; Telepolis, 11.2026)
- Government Plenipotentiary for Cybersecurity (Poland) – “Rekomendacja nr 2026-67a-10 dla sektora bankowości i infrastruktury rynków finansowych” (gov.pl, 17.11.2026)
- wnp.pl – the “bank employee” scam (12.2025)
- Antyweb (09.2024, 04.2026), Spidersweb (03.2023, 11.2024), rp.pl (11.2023), bank.pl (10.2026), Cashless (07.2023, 10.2026), ITHardware (07.2026), Revolut Help – sign-in methods in Polish services