I'm blocking connections from AWS to my on-prem services

Motivations for Blocking Cloud / AWS IPs

  • Main motivation: reduce “meaningless, low‑value, non‑human” traffic (scanners, bots, scrapers, spoofed pings) and avoid contributing to model training.
  • Author sees the blog and on‑prem services as for humans or known contacts; often shares links directly instead of relying on public discoverability.
  • Frustration with cloud providers’ abuse handling (perceived as unhelpful / one‑way) and with expectations of “just scale” to absorb abusive traffic.
  • For some, blocking clouds or specific ASNs is a pragmatic response to repeated abusive activity (DoS, DNS amplification attempts, spam, bad crawlers).

Critiques and Concerns About Blocking

  • Risk of losing search engine indexing, archive inclusion, and incidental discovery (including by smaller alternative search engines).
  • Possibility of collateral damage: blocking cloud IPs can also block:
    • VPN exit nodes for individuals and enterprises.
    • Desktop / Workspaces‑type services running in clouds.
  • Blocking large ranges is seen by some as contributing to Internet “balkanization”; they argue the blockers are themselves driving fragmentation.
  • Others argue it’s mostly symbolic: big AI and clouds can buy or obtain data indirectly anyway.

Data Center IP Reputation & Self‑Hosting

  • Several comments note that “data center IPs” are already second‑class:
    • Completely blocked by some streaming and copyright‑sensitive services.
    • Treated as suspicious by CDNs (more CAPTCHAs, stricter rate limits).
  • A rough hierarchy is described: residential > mobile > institutional > public cloud/hosting, with clouds being the most heavily filtered.
  • This trend is seen as an increasing barrier to self‑hosting.

Alternative Mitigations & Operational Practices

  • Suggested alternatives to blanket blocking:
    • Rate‑limiting (ICMP, HTTP, etc.).
    • Better TLS configuration, redirects, HSTS.
    • Using abuse contacts for specific incidents (though experiences with responsiveness vary).
  • Some admins accept background scans as “Internet radiation” and instead:
    • Harden services, require keys for SSH, VPN‑gate access, or bastion hosts only.
    • Use IPv6, WireGuard, OpenVPN, port knocking, or “no‑cat” style access gating.
  • There is debate over whether the real issue is technical (traffic volume) or social/legal (data usage, AI training, commercialization).