Prompt
How do I detect bot-driven fake account creation at signup?
Latest observation
To detect bot-driven fake account creation at signup, use a layered approach: combine behavioral signals, device/network signals, email/phone verification, and risk scoring. No single check is enough.
1) Look for behavioral patterns
Bots often behave differently from humans:
- Signup speed: extremely fast form completion
- Field timing: every field entered at the same pace, or instant submit
- Mouse/touch behavior: no movement, perfectly linear movement, or no focus changes
- Keystroke patterns: unnatural typing cadence, paste-heavy input
- Navigation sequence: direct hit to signup endpoint without normal browsing flow
- Repetition: same username/email patterns, same password structure, same referral source
2) Check device and browser signals
Collect lightweight fingerprinting signals and look for anomalies:
- User agent consistency
- Missing or spoofed browser APIs
- Headless browser indicators
- Canvas/WebGL/audio fingerprints
- Timezone/language mismatch
- Disabled cookies/localStorage
- Reused device fingerprint across many signups
Be careful: use these for fraud detection, not as the sole gate, because fingerprints can be noisy and privacy-sensitive.
3) Analyze IP and network reputation
Bots often come from suspicious infrastructure:
- Datacenter IPs
- VPN/proxy/Tor usage
- Known bad IP reputation
- Unusual geolocation relative to signup behavior
- Many accounts from the same IP, subnet, ASN, or device
Signals to track:
- IP velocity: accounts per IP per hour/day
- Subnet velocity
- ASN reputation
- Geo/IP mismatch
- IP churn patterns
4) Verify contact methods
A lot of fake signups use low-trust or disposable identifiers:
- Email domain reputation
- Disposable/temporary email detection
- MX record presence
- Phone number verification
- VOIP/virtual number detection
- Reuse of same email/phone across multiple accounts
5) Use challenge-response selectively
Don’t CAPTCHA everyone. Trigger it when risk is elevated:
- After suspicious behavior
- On high-volume IPs
- For referral abuse
- For signups with poor device/network reputation
Options:
- CAPTCHA / hCaptcha / Turnstile
- Email OTP
- SMS OTP
- WebAuthn for higher-trust workflows
6) Build a risk score
Combine signals into a score instead of hard-blocking on one indicator.
Example features:
- IP reputation score
- Device reuse count
- Signup velocity
- Email age / disposable domain flag
- Typing entropy / interaction score
- Browser integrity checks
- Referral abuse indicators
Then classify:
- Low risk: allow
- Medium risk: step-up verification
- High risk: block or queue for review
7) Watch for abuse after signup too
Bots may look normal at signup and act later:
- No profile completion
- No email verification
- Immediate spammy actions
- Same content posted across accounts
- Burst activity shortly after creation
Track:
- account age vs. first action
- first 10 actions
- content similarity across new accounts
- connection/follow/invite abuse
8) Use honeypots and hidden traps
Add non-visible form fields or decoy actions:
- Hidden fields that humans won’t fill
- Time-based validation on form completion
- JavaScript-required tokens
- Fields rendered only to bots via naive scraping
Use carefully so you don’t harm accessibility or legitimate users.
9) Correlate across entities
Bots often create clusters:
- Same IP + same device + similar email pattern
- Shared referral codes
- Repeated signup times
- Same phone carrier/region pattern
- Similar browser fingerprints
Graph analysis can help detect coordinated signups.
10) Continuously tune and monitor
Measure:
- False positives
- Blocked bot rate
- Conversion impact
- Appeal/recovery rate
- Abuse recidivism
Update rules as attackers adapt.
Practical implementation pattern
A common signup flow:
- User submits signup form
- Backend computes a risk score from:
- IP reputation
- device fingerprint
- email/phone trust
- behavior/timing
- velocity / cluster history
- If low risk → create account
- If medium risk → require CAPTCHA / email OTP
- If high risk → block, rate limit, or queue for review
Good starting signals
If you want the fastest wins, start with:
- IP/device velocity checks
- Disposable email detection
- CAPTCHA only for suspicious signups
- Email verification
- Rate limiting by IP/subnet/device
- Behavioral timing checks
Important caution
Avoid relying too much on aggressive fingerprinting alone. It can:
- hurt privacy
- create false positives
- break for users behind shared networks, VPNs, or privacy tools
A combined, risk-based system is much more robust.
If you want, I can also help you design a signup fraud detection schema, a feature list for a risk model, or example rules/SQL for detecting bot signups.