Phishing: how it works
How a phishing attack unfolds, from the decoy message to stolen credentials, and which reflexes an employee should adopt.
On this page
In 30 seconds
- Where it happens
- By email, text message, direct message on a social network or phone call[2].
- What the attacker wants
- To make you open a link or an attachment, or to make you give your credentials[2].
- The reflex
- Do not click, do not call back the number in the message, and check with the supposed sender by a means that you choose yourself[2].
- If it is already done
- Change the password concerned, keep the message and report it[1].
What it is
Phishing (hameçonnage in French) is a fraudulent technique that aims to push the victim into handing over personal data, such as access accounts and passwords, or banking data, by posing as a trusted third party[1]. The message can appear to come from a bank, a social network, a carrier, an online shop or a government agency[1].
For an employee, the stakes go beyond a personal account. The goal is to steal personal or professional information[1]. A company login obtained through a fake message can open the door to the network.
Straight answers
What distinguishes targeted phishing from mass phishing?
All forms of phishing are social engineering delivered by electronic means. When the attack targets a specific person, company or sector, it is called spearphishing. Without a specific target, it is for example a spam campaign distributing malware[3].
Is a well-written message necessarily safe?
No. According to CISA, grammar mistakes used to be a common sign, but in the age of artificial intelligence some emails have perfect grammar and spelling. The other signs must therefore be looked for[2]. Cybermalveillance.gouv.fr likewise notes that the attempts are better and better executed[1].
Does the attack only come through email?
No. MITRE describes emails with an attachment or a link, messages through third-party services (social networks, for example) and calendar invitations[3]. A victim can also receive a message asking them to call a number, where they are directed to a malicious address, made to download malware or to install a remote management tool[3].
Inside the attacker's head
The attacker first chooses a plausible pretext. The text of a booby-trapped email generally gives a good reason to open the file, and may even explain how to get around the protections on the computer[4].
The attacker can also win the target's trust before sending anything. MITRE describes the creation of fake social network accounts to offer job opportunities to employees, which gives a reason to ask questions about the services and software in place[5]. The malicious content is then sent to a personal mailbox that the victim checks from a work computer, which bypasses some restrictions of the corporate mail system[5]. Another method is to slip into an existing conversation. The attacker adds the target to an email thread that already contains malicious files or links ("thread hijacking")[3].
What you see on screen
The signs described by CISA are: urgent or emotional language that announces serious consequences if you do not respond right away, a request for personal or financial information, untrustworthy shortened addresses, and incorrect addresses or links, such as "amazan.com"[2].
CISA also tells the story of a busy professional who is shopping online and receives an email announcing a failed payment and inviting them to log in. The link led to a fake seller site. The hackers stole their password and card data[2].
Cybermalveillance.gouv.fr adds other signals: a tempting offer, a suspicious look, an unexpected attachment, a fanciful sender address[1].
Under the hoodIT teams
Three mechanisms recur.
- The attachment: Office document, executable, PDF or archive. When opened, the content exploits a flaw or runs directly. Attackers often change extensions and icons so that an executable looks like a document[4]. The email can also supply the password of an archive to escape filters[4].
- The link: it downloads the malware instead of attaching it, which avoids the checks on attachments. The site visited can exploit the browser, or offer files to download[6].
- Spoofing: the sender can be forged (Email Spoofing), which fools the recipient and also the automatic security tools[3].
Lookalike addresses take several forms. MITRE cites the IDN homograph attack, where special characters imitate a legitimate site, and addresses scrambled by taking advantage of quirks of the format, such as ignoring everything that comes before the "@" symbol[6]. Neutralized illustration taken from MITRE:
Illustration, neutralized address:
hxxp://google.com@1157586937
[the browser ignores "google.com" and goes to the numeric address]Cybermalveillance.gouv.fr points out that a single character can change in a site's address to deceive[1]. The link can also ask for permissions: in "device code phishing", the victim is led to enter a code on an authorization page, which produces a token for the attacker[6].
Case filesReal cases
Midnight Blizzard on Microsoft Teams (since late May 2023)
On 2 August 2023, Microsoft describes a campaign of Teams messages run by Midnight Blizzard, an actor based in Russia, attributed by the US and UK governments to the Russian Foreign Intelligence Service (SVR)[7]. The observed pattern dates from at least late May 2023[7]. The attacker uses Microsoft 365 environments (tenants) of small businesses already compromised, creates security-themed subdomains and sends the messages from them[7].
The sequence has three steps: a chat request from an external user posing as technical support or the security team, then a message that pushes the target to enter a code in the Microsoft Authenticator app, then, if they do so, a token that lets the attacker authenticate as them[7].
Illustration (paraphrase, neutralized address):
From: support-securite[.]exemple[.]invalid ("external" user)
[message claiming to come from support, asking to enter
the displayed code in the authentication app]According to Microsoft, fewer than 40 organizations worldwide were affected[7]. Microsoft recommends never entering an authentication code received through an unsolicited message, checking the "External" label and never allowing a sign-in through a messaging app[7].
Scattered Spider (advisory of 16 November 2023, updated 29 July 2025)
The FBI, CISA and Canadian, Australian and British agencies published a joint advisory on Scattered Spider, a cybercriminal group that targets large companies and their outsourced IT help desks[8]. The group posed as IT support by phone call or text message to obtain credentials, asked employees to launch remote access tools, and asked them for their one-time code[8]. It also sent repeated authentication notifications until the employee pressed "Accept" (MFA fatigue)[8]. The July 2025 update adds the reverse case: posing as an employee to the help desk to get their password reset[8].
What you do
- Never give out sensitive information by message or phone: no serious government agency or company asks for banking data or passwords through these channels[1].
- Hover over a link without clicking to see the real address, or use a bookmark that you created[1].
- Check the address displayed in the browser. At the slightest doubt, close the page[1].
- If the message seems real, do not call back the number it contains: look up the contact details on the official site, or reach the person another way[2].
- Do not delete before reporting: look for the report button near the sender's address, then delete the message without clicking, including on "unsubscribe"[2].
- In France, report an email to Signal Spam, a text message to 33700 and a website to Phishing Initiative[1].
If you have given out a password, change it on the service concerned and on all those where you used it. Keep the message as evidence. If you gave out banking details, request a block on the card[1].
Reporting has an effect beyond your own case. The UK NCSC has the power to investigate scam addresses and sites and to have them taken down, and reporting reduces the number of messages received and protects other people[9]. In September 2026, it had received more than 59.6 million reports, which had led to the takedown of 263,700 scams across 466,300 addresses[9]. The information you publish online, notably on social networks, makes messages more convincing: reviewing your privacy settings reduces the risk[9].
Sources
- Cybermalveillance.gouv.fr, “Que faire en cas de phishing ou hameçonnage ?”, 10 January 2020. https://www.cybermalveillance.gouv.fr/tous-nos-contenus/fiches-reflexes/hameconnage-phishing (accessed 10 November 2026)
- CISA, “Recognize and Report Phishing”. https://www.cisa.gov/secure-our-world/recognize-and-report-phishing (accessed 10 November 2026)
- MITRE ATT&CK, “Phishing, Technique T1566 - Enterprise | MITRE ATT&CK®”. https://attack.mitre.org/techniques/T1566/ (accessed 10 November 2026)
- MITRE ATT&CK, “Phishing: Spearphishing Attachment, Sub-technique T1566.001 - Enterprise | MITRE ATT&CK®”. https://attack.mitre.org/techniques/T1566/001/ (accessed 10 November 2026)
- MITRE ATT&CK, “Phishing: Spearphishing via Service, Sub-technique T1566.003 - Enterprise | MITRE ATT&CK®”. https://attack.mitre.org/techniques/T1566/003/ (accessed 10 November 2026)
- MITRE ATT&CK, “Phishing: Spearphishing Link, Sub-technique T1566.002 - Enterprise | MITRE ATT&CK®”. https://attack.mitre.org/techniques/T1566/002/ (accessed 10 November 2026)
- Microsoft, “Midnight Blizzard conducts targeted social engineering over Microsoft Teams”, 2 August 2023. https://www.microsoft.com/en-us/security/blog/2023/08/02/midnight-blizzard-conducts-targeted-social-engineering-over-microsoft-teams/ (accessed 10 November 2026)
- CISA, “Scattered Spider”, 16 November 2023. https://www.cisa.gov/news-events/cybersecurity-advisories/aa23-320a (accessed 10 November 2026)
- NCSC (UK), “Phishing scams: how to spot and report them”, 26 November 2021. https://www.ncsc.gov.uk/collection/phishing-scams (accessed 10 November 2026)
Cite this article
Brain, “Phishing: how it works”, Cyberpedia, 2026. https:/
BibTeX
@misc{brain_cp001_en_2026,
author = {Brain},
title = {Phishing: how it works},
year = {2026},
howpublished = {Cyberpedia},
url = {https://startbrain.io/cyberpedia/en/articles/phishing-mechanics-vectors/},
note = {CC BY 4.0}
}