Default Rule Packs

Packs shipped with Sentry, enabled with a single line. Each pack is a set of rules with tags for easy inspection/editing via CLI.

Pack	Default	What it does
vpn_proxy	on	Block/Challenge IPs classified as VPN/proxy (reputation = Vpn/Proxy)
tor	on	Block Tor exit nodes (reputation = Tor)
datacenter_abuse	on	Challenge datacenter ASNs not in allowlist (DigitalOcean, OVH, Hetzner, etc. — bot targets)
crawlers_bad	on	Block UAs of scanners/attack tools: sqlmap, nikto, nmap, masscan, zgrab, suspicious curl/8.*, python-requests with no context
crawlers_good	off	Allow legitimate bots (Googlebot, Bingbot, etc.) — verified via reverse-DNS per Google's spec
empty_ua	on	Challenge/block requests with no User-Agent (rare in legitimate traffic)
sensitive_paths	on (enforce)	Block hits on sensitive files/dirs by default (see full list below)
country_blocklist	off	Block unsupported countries (configure ISO list)
country_allowlist	off	Allow only listed countries (more restrictive, opt-in mode)
http_anomaly	on	Block rare unused methods (TRACE, CONNECT), HTTP/0.9, malformed headers
rate_scan	on	Rate-limit/Block IP with >N 404s in a window (directory brute-force)

Default on semantics: packs ship active but in Log or Challenge mode (not direct Block) on first deploy — shadow mode to validate before hardening. The user promotes to Block after confirming zero false positives. Controlled by mode = "shadow" | "enforce" per pack. Exception: sensitive_paths ships in enforce by default (access to .env/.git is always malicious).

sensitive_paths pack — full list (default enforce)

Files and directories whose access is always blocked by default. Coverage is split into categories; each entry is a path regex → Block rule. The list is extensible via config/DB.

Credentials & configuration

\.env(\.local|\.production|\.development)?$      # .env, .env.local, ...
\.env\.[a-z]+$                                    # any .env.* variant
config\.(php|json|yml|yaml|ini|conf)              # app configs
secrets\.(json|yml|yaml)
credentials\.(json|csv)
\.htpasswd
wp-config\.php
local\.xml                                        # Magento
settings\.php                                     # Drupal
configuration\.php                                # Joomla

SCM & directory metadata

/\.git/                                           # .git/, HEAD, config, index
/\.svn/
/\.hg/
/\.bzr/
/\.gitignore
/\.gitattributes
/\.dockerignore

Cloud & infrastructure

/\.aws/                                           # credentials, config
/\.ssh/                                           # id_rsa, id_ed25519, authorized_keys
/\.gcp/
/\.azure/
/\.kube/                                          # kubeconfig
/\.docker/                                        # config.json with registry tokens
/\.terraform(\.tfstate)?

Build files & artifacts

/(package-lock\.json|yarn\.lock|composer\.lock)   # optional: version info for recon
/(docker-compose\.yml|docker-compose\.yaml)       # exposes service topology
/(Dockerfile|Puppetfile|Vagrantfile)
/\.npmrc                                          # npm tokens
/\.pypirc                                         # pypi tokens
/\.netrc                                          # HTTP creds

Admin panels & known tools

/(wp-admin|wp-login\.php)                         # WordPress
/(phpmyadmin|pma|phpMyAdmin)                      # phpMyAdmin
/(adminer|adminer\.php)
/(wp-content/uploads/phpmailer)                   # common exploit
/manager/                                         # Tomcat manager
/server-status                                    # Apache mod_status
/server-info
/nginx-status
/fpm-status
/actuator(/env|/heapdump|/threaddump)?            # sensitive Spring Boot actuator
/health(/.*)?                                     # optional (may be legit)

Backup & dump

\.(sql|bak|backup|old|swp|tmp|orig|save|copy)$
/(dump|backup|db)\.(sql|tar|gz|zip|tgz)
/www\.(zip|tar|gz|rar|7z)                         # full-site dumps

System & dangerous exposures

/\.well-known/security\.txt$        # ALLOW (legitimate — RFC 9116) → explicit allowlist
/\.DS_Store
/Thumbs\.db
/(etc/passwd|etc/shadow)             # path traversal via decode
/(proc/self/environ|proc/self/fd/.*)

Technical implementation

• Each category is an individually toggleable sub-pack (sentry rules packs list shows granular state).
• The internal allowlist always permits /.well-known/security.txt (RFC 9116 — public responsible disclosure document) even with the pack active.
• Case-insensitive match (.ENV == .env) to avoid trivial bypass.
• Considers encodings: %2e (.), %2f (/), ..;/ (path traversal smuggling), double-encoding — pre-match normalization.
• Routes explicitly allowlisted by the user ([[rules.custom]] action = "allow") take priority over the pack, allowing /admin/ to be exposed if the app genuinely needs it.

Why enforce and not shadow from the start

Accesses to .git/, .env, .ssh/ are statistically 100% malicious in web apps (there's no legitimate reason for a browser to access these). The cost of a false positive here is zero versus the risk of leaking credentials.

Config in sentry.toml

[[rules.pack]]
name = "sensitive_paths"
mode  = "enforce"

[[rules.pack]]
name = "vpn_proxy"
mode  = "shadow"

[[rules.pack]]
name = "tor"
mode  = "enforce"

[[rules.pack]]
name = "crawlers_bad"
mode  = "enforce"

[[rules.pack]]
name = "crawlers_good"
mode  = "enforce"         # allow Googlebot etc.

[[rules.pack]]
name = "country_blocklist"
mode  = "enforce"
countries = ["RU","CN","KP"]   # ISO codes