Detecting AI-Driven Attacks: Deepfakes, Fake Recruiters, Cloned Sites
AI makes three attack types more convincing: deepfake voice and video used to impersonate executives or relatives, fake recruiters and fake job candidates used to steal data or gain insider access, and cloned websites that copy real brands to steal logins and payments.
Topic: AI in Cybersecurity in Cybersecurity & Ethical Hacking · Explainer
Also covers: Cyber Threats, Scams & Online Safety
Quick answer: AI makes three attack types more convincing: deepfake voice and video used to impersonate executives or relatives, fake recruiters and fake job candidates used to steal data or gain insider access, and cloned websites that copy real brands to steal logins and payments. The most reliable defences are processes rather than detection tools: verify any payment or access request through a separate known channel, verify identities rigorously during hiring, and monitor for lookalike domains and certificates while using phishing-resistant MFA that fake sites cannot capture.
Key takeaways
- Deepfake detection by eye is unreliable; verification processes work better.
- The FBI has warned about fraudulent remote IT workers using false identities.
- Fake recruiters often move candidates to messaging apps and send malicious files or tests.
- Certificate transparency logs reveal lookalike domains early.
- Passkeys defeat credential theft on cloned sites.
Table of Contents
- Which AI-Driven Attacks Are Most Common?
- How Do You Stop Deepfake Fraud?
- How Do You Spot Fake Recruiters and Fake Candidates?
- How Do You Detect Cloned Websites?
- Which Controls Work Across All Three?
- What Does a Response Playbook Look Like?
- Which Hiring Checks Stop Fraudulent Remote Workers?
- References
Which AI-Driven Attacks Are Most Common?
| Attack | What AI adds | Key defence |
|---|---|---|
| Deepfake voice and video | Realistic impersonation of trusted people | Out-of-band verification, payment controls |
| Fake recruiters | Polished profiles and messages at scale | Verify company and recruiter, never run sent files |
| Fake job candidates | Generated resumes, face swaps in interviews | Strong identity checks, in-person or verified ID |
| Cloned websites | Fast, accurate copies of real sites | Domain monitoring, passkeys, takedown requests |
The FBI's December 2024 PSA describes criminal use of generative AI for fraud.
How Do You Stop Deepfake Fraud?
- Require call-back on a known number for any payment, bank change or credential request.
- Use dual approval for payments above a threshold.
- Agree code words for sensitive requests.
- Train staff with real examples from the NSA, FBI and CISA deepfake guidance.
How Do You Spot Fake Recruiters and Fake Candidates?
For job seekers:
- Check the job on the company's official careers page.
- Be wary of recruiters who move quickly to WhatsApp or Telegram, ask for fees, or send "coding tests" to run locally; some contain malware.
- Never share bank details or ID scans before a verified offer.
For employers: the FBI has warned that North Korean IT workers use false or stolen identities to get remote jobs and access company systems (IC3 PSA).
- Verify identity documents through trusted services; hold at least one in-person or live verified meeting.
- Watch for mismatches: camera reluctance, inconsistent locations, laptops shipped to unexpected addresses.
- Apply least privilege to new hires and monitor remote access tools.
How Do You Detect Cloned Websites?
New HTTPS sites need certificates, which are published in Certificate Transparency logs. Monitoring them reveals lookalike domains early:
# Certificates issued for names containing your brand (replace examplebank)
curl -s "https://crt.sh/?q=%25examplebank%25&output=json" |
jq -r '.[].name_value' | sort -u | grep -v "examplebank.com$"
Use crt.sh or a monitoring service, register common misspellings of your domain, and request takedowns from hosting providers and registrars. For users, passkeys only work on the genuine domain, so cloned sites cannot steal them.
Which Controls Work Across All Three?
- Phishing-resistant MFA for staff and customers.
- Verification procedures that never rely on a single channel.
- Awareness training updated with AI-enabled examples, per CISA's phishing guidance.
- Easy reporting routes and quick response.
- In India, report fraud on 1930 or cybercrime.gov.in.
WebAsha offers security awareness training.
What Does a Response Playbook Look Like?
- Pause the transaction or access request.
- Verify through a known channel: call back, in-person check, or a separate messaging system.
- If fraud is suspected, preserve evidence: recordings, messages, email headers.
- Report internally to security and finance; contact the bank if money moved.
- Report externally: in India, 1930 and cybercrime.gov.in; and CERT-In for organisational incidents.
- Share lessons with staff.
Align the playbook with NIST SP 800-61 Rev. 3 incident response recommendations.
Which Hiring Checks Stop Fraudulent Remote Workers?
| Check | Purpose |
|---|---|
| Live video interview with cameras on throughout | Reduces proxy interviewing |
| Identity verification against official documents | Detects stolen or fake identities |
| Reference checks via independently found contacts | Avoids fake referees |
| Ship equipment only to verified addresses | Prevents laptop farms |
| Monitor for remote access tools and unusual logins | Detects misuse after hiring |
The US Department of Justice has announced actions against schemes placing North Korean IT workers in companies; employers should combine HR and security checks.
References
- FBI IC3, Generative AI financial fraud PSA
- FBI IC3, North Korean IT worker PSA
- NSA, FBI and CISA, Contextualizing Deepfake Threats
- crt.sh Certificate Transparency search
- CISA and partners, Phishing Guidance
- FIDO Alliance, Passkeys
- NIST SP 800-61 Rev. 3
- US Department of Justice, Office of Public Affairs
Frequently Asked Questions
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Wow
0
Sad
0
Angry
0