OnlyFake: A site where ‘neural networks’ churn out fake IDs

Original Article ↗ Hacker News Discussion ↗

AI-generated fake IDs are becoming cheap and convincing enough to bypass many online “know your customer” (KYC) checks, underscoring how weak photo-based identity verification has always been. Commenters argue this will accelerate the need for stronger, government- or bank-backed digital identity systems (often based on public‑key cryptography), while highlighting the political, security, and privacy obstacles to deploying such infrastructure, especially in the US. Many conclude that remote identity proof based on static images or simple selfies is rapidly becoming obsolete in the face of improving generative tools.

Access Model and Anti-Scraping Measures

  • Several comments criticize 404 Media’s email wall, noting odd behavior where only part of a paragraph is hidden and calling it annoying “email mining.”
  • Others link to 404 Media’s explanation that the email requirement is to reduce AI scraping and content republishing.

Weakness of Photo-Based ID Verification

  • Many argue that relying on a static photo of an ID for identity verification or KYC is fundamentally flawed and was always insecure.
  • Some see OnlyFake as a useful stress test that will force institutions to abandon “send a picture of your ID” flows.
  • Others note that even “selfie holding ID” or “turn your head” flows are likely to become trivial to fake with video deepfakes.

Neural Networks vs Traditional Fake-ID Generation

  • Some say this doesn’t sound technically hard, and older techniques (ImageMagick, texture synthesis, Photoshop) already existed.
  • Others respond that neural networks add realistic randomness (lighting, reflections, textures) that defeats pattern-based detectors trained on millions of real IDs.
  • There’s consensus that high-skill attacks quickly become products in the criminal ecosystem.

KYC, Banks, and Crypto

  • Core concern: bypassing KYC for banks, gambling, and crypto exchanges, not physical fake IDs for bars.
  • Some claim serious services already use stronger checks (live video, face-match, third‑party KYC APIs), though others say many still accept weak document photos.
  • Discussion highlights that in some jurisdictions there is limited or patchy API access to driver’s-license databases, so services fall back to document images.
  • Several point out that if photo fakes succeed, those KYC providers are not doing real verification.

Alternative Digital ID Systems

  • Many propose PKI-based or smartcard-based identity: government CAs, NFC-enabled IDs, or phone-based secure elements.
  • Examples cited from the thread: Estonia-style e-ID, EU’s eIDAS and Digital Identity Wallet, national smartcards, and bank‑run ID systems like BankID (Sweden, Czech Republic), DigiD (Netherlands), Diia (Ukraine), and US federal smartcards plus id.me/login.gov.

Centralization, Trust, and Politics

  • One side argues centralized, cryptographic ID is clearly safer than today’s “cardboard box” of screenshots and SSNs.
  • Others warn a single global or national identity system concentrates risk: breaking it could endanger “all the money” or enable oppression, citing historical abuses of national IDs.
  • US-specific debate centers on federal vs state control, constitutional limits, religious and ideological resistance, and the de facto use of SSNs as weak national identifiers.

Elections and Fraud

  • Questions about voter fraud are raised; multiple replies say AI fake IDs are largely irrelevant to US in‑person voting, which depends on pre‑registration and precinct lists.
  • Most see financial fraud, money laundering, and account takeovers as the main threat vector, not voting.