Security and Safe APIs

Unpickling is not parsing. A pickle stream is a tiny program whose opcodes import names and call them, so pickle.loads on attacker-controlled bytes can invoke any importable callable with attacker-chosen arguments before it returns anything.

Placeholders keep the statement and the data on separate channels. sqlite3 compiles the SQL text first, then binds each parameter into a slot of the compiled statement, so a value can never become syntax. String formatting merges the two.

The data filter refuses tar members that would land outside the destination directory, refuses links whose target is absolute or escapes, and rejects device and FIFO members, while clearing ownership and risky permission bits. Python 3.14 makes it the default for extraction.

Python's format mini-language allows attribute access and indexing inside a placeholder, so whoever writes the template chooses what is read out of the objects you pass. A chain through a method's globals reaches module-level secrets.

Python's re module uses a backtracking engine. Nesting one quantifier inside another over the same characters gives it exponentially many ways to split the input, so a 30-character string that ultimately fails to match costs about a billion steps.

eval() compiles and runs its text as a Python expression, so one expression can import a module and touch files, processes or sockets with the privileges of your process. Parse untrusted text; never evaluate it.

The Python object graph is fully connected: from any value you can follow attributes back to the type system, the builtins, and every loaded module. No matter which names you hide, code can walk to them. Real isolation is an OS-level boundary — a separate process, container, or VM.

Importing a module executes that module's top-level code. If the caller names the module, the caller chooses which code runs, and search order plus every installed distribution decide what that name resolves to. Import from a fixed allowlist instead.

ast.literal_eval evaluates only literal structures: strings, bytes, numbers, booleans, None, Ellipsis, and the tuple, list, dict and set displays built from them. Any name, attribute, operator or function call raises ValueError instead of running.

Every object exposes class, and from there bases reaches object, whose subclasses() lists every class the interpreter has already loaded. An attacker scans that list for a gadget class whose method or globals reaches os or subprocess, so stripping builtins never removes the path back to dangerous code.

The process that launched yours supplies the environment block, and Python copies it into os.environ when the os module is first imported. Whoever controls the launch - a shell, a container spec, a job scheduler - controls the values.

With a list, CPython execs the program directly and each element becomes exactly one argv entry, so nothing tokenises it. With shell=True the whole string goes to /bin/sh -c, where ; | $() and word splitting are live syntax.

Python puts the script's own directory first on sys.path, ahead of the standard library. Anyone who can write there can drop a file named like a module the script imports, and their code runs at import time.

os.path.join throws away everything to the left of a component that is already absolute, so joining '/etc/passwd' onto a base returns '/etc/passwd'. A relative name holding '..' also climbs out. Joining concatenates; it never confines.

tempfile.mktemp only returns a name. Between that return and your own open() an attacker can create that path, usually as a symlink. tempfile.mkstemp creates and opens the file itself in one exclusive step, so there is no window to exploit.

A hashlib hash object accumulates bytes across repeated update() calls, so feeding data in chunks matches hashing it in one call. digest() returns the raw bytes, hexdigest() the same value as a hex string; reading either does not reset the object.

hashlib offers two password functions: pbkdf2_hmac and scrypt. Both take a per-user salt and a work factor you choose, so one guess costs real time. hashlib.sha256 is built to be fast, which is exactly wrong for passwords.

The random module's default generator is a Mersenne Twister whose entire internal state can be reconstructed from a few hundred observed outputs, so later tokens become predictable. secrets draws each token from the operating system's cryptographic source instead.

== on two bytes objects stops at the first differing byte, so the time it takes reveals how much of a secret an attacker guessed right. hmac.compare_digest always does the same work; secrets.compare_digest is the same function.

It turns off certificate verification, so a Python client accepts any certificate at all, including one an attacker generated. The connection stays encrypted but is no longer authenticated, which defeats the point of TLS.

A formatted traceback exposes absolute file paths, the module layout, the failing source lines and sometimes the repr of local values such as connection strings. Log it server side under an opaque reference id and return only that id with a generic message.

The logging module interpolates %s arguments only when a record is actually emitted, so a filtered-out call costs nothing and the raw values stay on LogRecord.args, where a redaction filter can still rewrite them. An f-string bakes them in first.

Declare the attribute with dataclasses.field(repr=False). The generated repr then omits it, and str() omits it too because it falls back to repr. The value is still an ordinary attribute, so dataclasses.asdict(), vars() and pickling still carry it.

The -O flag sets the builtin __debug__ to False and makes the compiler omit every assert statement from the bytecode entirely. A check written as an assert then guards nothing, so an assert must never carry input validation or a security decision.

getpass.getpass() turns terminal echo off while the user types, so the password never appears on screen or in a terminal recording, and it reads from the controlling terminal rather than a redirected stdin. input() echoes every character.