Security Essentials

Security Essentials

Use this skill before writing ANY security-sensitive code.

Canonical FP bar: docs/fcis-engineering-rules.md — Functional Core, Imperative Shell: pure domain modules; side effects at edges. Parse/allowlist at the boundary; never trust raw maps deep in core.

RULES — Follow these with no exceptions

1. Never call String.to_atom/1 on user input — use String.to_existing_atom/1 or a whitelist case (Atom Table Exhaustion) 2. Never interpolate user input into an Ecto fragment or raw SQL — use ^value bindings, field/2, or $1/$2 placeholders (SQL Injection) 3. Never redirect to a user-controlled URL — use ~p"..." verified routes or a whitelist of allowed paths (Open Redirects) 4. Never render user content with raw/1 in HEEx — let auto-escaping run, or sanitize with HtmlSanitizeEx first (Cross-Site Scripting (XSS)) 5. Never log secrets — log identifiers (user_id, email), never passwords, tokens, or PII (Sensitive Data in Logs) 6. Always compare tokens with Plug.Crypto.secure_compare/2 — never ==, which leaks length/content via timing (Timing Attacks) 7. Never disable Phoenix CSRF protection — keep :protect_from_forgery in the browser pipeline and use <.form>, not raw <form> (CSRF Protection) 8. Always authorize before returning a record — verify ownership to prevent parameter tampering and IDOR (Common Vulnerable Patterns) 9. Always run mix deps.audit && mix hex.audit && mix sobelow before merge — fail CI on any HIGH or CRITICAL finding (Dependency Auditing)

See assets/security_checklist.md for a copy-paste pre-merge checklist covering input validation, auth, secrets, and production hardening.

Security Review Process

Apply this sequence whenever writing or reviewing security-sensitive code:

1. Identify attack surface — list all user inputs, external data sources, and network boundaries
2. Apply RULES — implement the feature following all rules above
3. Run mix sobelow --router MyAppWeb.Router — static analysis on your router and controllers
4. Run mix sobelow --private — check private functions for vulnerabilities
5. Review findings by severity — HIGH severity first, then MEDIUM, then LOW
6. Fix each finding — apply the correct pattern from the sections below
7. Re-run sobelow — repeat until no issues reported
8. Run full audit — mix deps.audit && mix hex.audit && mix sobelow before merging
9. Test manually — verify with curl/introspection that expected inputs are rejected

Common Vulnerable Patterns

Parameter Tampering

❌ Bad — trusting user input:

def index(conn, %{"status" => status}) do
  users = Repo.all(from u in User, where: u.status == ^status)
  render(conn, "index.html", users: users)
end

✅ Good — validate against allowed values:

@allowedStatuses ~w(active inactive pending)
def index(conn, %{"status" => status}) do
  if status in @allowedStatuses do
    users = Repo.all(from u in User, where: u.status == ^status)
    render(conn, "index.html", users: users)
  else
    put_status(conn, :bad_request)
    |> json(%{error: "Invalid status"})
  end
end

IDOR (Insecure Direct Object Reference)

❌ Bad — no authorization check:

def show(conn, %{"id" => id}) do
  user = Repo.get!(User, id)
  render(conn, "show.html", user: user)
end

✅ Good — verify ownership:

def show(conn, %{"id" => id}) do
  current_user = conn.assigns.current_user

  case Accounts.get_user_for_current_user(current_user, id) do
    {:ok, user} -> render(conn, "show.html", user: user)
    {:error, :not_found} -> put_status(conn, :not_found) |> json(%{error: "Not found"})
    {:error, :unauthorized} -> put_status(conn, :forbidden) |> json(%{error: "Forbidden"})
  end
end

Atom Table Exhaustion

❌ Bad — user controls the atom:

role = String.to_atom(params["role"])

✅ Good — whitelist approach:

case params["role"] do
  "admin" -> :admin
  "user" -> :user
  "moderator" -> :moderator
  _ -> {:error, :invalid_role}
end

SQL Injection

Never interpolate strings into Ecto queries; use ^variable or $1/$2 placeholders.

❌ Bad — string interpolation in fragment:

# NEVER do this — user can inject SQL through field or value
from(u in User, where: fragment("lower(#{field}) = ?", ^value))
from(u in User, where: fragment("#{condition}", []))

❌ Bad — using unvalidated input in raw SQL:

# NEVER do this — even with ~s() sigil
Ecto.Adapters.SQL.query(Repo, "SELECT * FROM users WHERE name = '#{name}'", [])

✅ Good — parameterized fragment with field/1:

# Safe — field is an atom from schema, value is parameterized
from(u in User, where: fragment("lower(?) = ?", field(u, :status), ^value))

✅ Good — parameterized raw SQL:

Ecto.Adapters.SQL.query(Repo, "SELECT * FROM users WHERE id = $1", [id])
Ecto.Adapters.SQL.query(Repo, "SELECT * FROM users WHERE name = $1 AND status = $2", [name, status])

✅ Good — Ecto query expressions always safe:

# Ecto query expressions are always parameterized
from(u in User, where: u.status == ^status and u.name == ^name)

Open Redirects

Never redirect to user-controlled URLs; use ~p"..." or a whitelist.

❌ Bad — user controls redirect destination:

def create(conn, %{"redirect_to" => redirect_to} = params) do
  redirect(conn, to: redirect_to)
end

✅ Good — use verified routes:

redirect(conn, to: ~p"/dashboard")

✅ Good — validate against known paths:

@allowed_redirects ["/dashboard", "/profile", "/settings"]

def create(conn, %{"redirect_to" => redirect_to} = params) do
  if redirect_to in @allowed_redirects do
    redirect(conn, to: redirect_to)
  else
    redirect(conn, to: ~p"/dashboard")
  end
end

Cross-Site Scripting (XSS)

Avoid raw/1; sanitize with HtmlSanitizeEx if HTML is required.

❌ Bad — bypasses escaping:

<%= raw(@user_bio) %>

✅ Good — let Phoenix auto-escape:

<%= @user_bio %>

✅ Good — sanitize if HTML rendering is required:

<%= raw(HtmlSanitizeEx.html5(@user_bio)) %>

Sensitive Data in Logs

❌ Bad:

Logger.info("User login", email: email, password: password)
Logger.debug("API call", token: api_token, response: resp)

✅ Good:

Logger.info("User login", email: email, user_id: user.id)
Logger.debug("API call", endpoint: url, status: resp.status)

Timing Attacks

Use Plug.Crypto.secure_compare/2 for token comparison; never ==.

❌ Bad — timing-unsafe:

def verify_token(provided_token, stored_token) do
  provided_token == stored_token
end

✅ Good — constant-time comparison:

def verify_token(provided_token, stored_token) do
  Plug.Crypto.secure_compare(provided_token, stored_token)
end

Dependency Auditing

# Check for known vulnerabilities in dependencies
mix deps.audit

# Verify package checksums against Hex
mix hex.audit

# Static security analysis on your code
mix sobelow --router MyAppWeb.Router

# Run all three before any merge
mix deps.audit && mix hex.audit && mix sobelow --config

Sobelow categories:

Add to CI pipeline:

# .github/workflows/security.yml
- name: Security Audit
  run: |
    mix deps.audit
    mix hex.audit
    mix sobelow --config

# mix.exs
defp aliases do
  [
    "security.check": ["deps.audit", "hex.audit", "sobelow --config"]
  ]
end

Interpretation: Any Sobelow finding of HIGH or CRITICAL severity MUST be fixed before merging. LOW findings should be tracked and addressed within 2 sprints.

CSRF Protection

Never disable Phoenix's built-in CSRF protection.

# Phoenix forms automatically include CSRF tokens
# <.form> component handles this — never use raw <form> tags

# API pipeline should NOT include :protect_from_forgery
pipeline :api do
  plug :accepts, ["json"]
  # No :protect_from_forgery — APIs use Bearer tokens instead
end

Common Pitfalls

Integration

Companion skills: phoenix-authorization-patterns, phoenix-auth-customization, phoenix-liveview-auth, code-quality.