Finding an authorization bypass on my own website
maxwelldulin.com
maxwelldulin.com
And still much preferable to not having it at all.
Don’t let perfect be the enemy of good.
The alternative for people using a library like this is to send plain queries to the db.
/s/today/30 years ago: SQL-92 describes them in chapter 4.18 - which at the time was just standardizing something almost all vendors already had in some form or another, for example, I'm pretty sure even the first release of ODBC in the late 1980s also had parameters.
Disclaimer: My pontifications below derive from my life's experiences starting with Access/JET Red as a sprog, to my current professional work with MS SQL Server, Postgres (and MySQL, I'll admit - but I'll say MariaDB instead) built over the past 17 years - but I have absolutely zero experience with Oracle and Db2, so I honestly don't know what cool language features they have (and they must be way ahead of the ISO spec, otherwise why else would people pay so much for it?... Hmm, then again, Oracle still doesn't have a bit/bool column type, does it?).
Anyway:
However, the expressiveness of parameters hasn't changed much since the original design of ODBCS - at least as far as I'm aware: Query parameters are still mere scalar value placeholders instead of hardcoding literals inside queries and statements: with limited exceptions like T-SQL's ceremony-laden table-valued-parameters and Postgre's array types, it's still not possible to parameterize database object identifiers or even have a true variadic `WHERE IN ( ... )` predicate without resorting to Dynamic-SQL, nor can we use parameters to conditionally enable or disable query predicate clauses: yet these are all essential features for any kind of ORM or language-integrated-query system built today (the canonical example being Linq/EF in .NET and TypeORM or Prisma for TypeScript/JS.
Why isn't anyone meaningfully advocating for a "SQL/2"? I know backwards-compatibility is 100% essential, but that's the easy part (because relational calculus and relational are isomorphisms, hence why semantic-preserving query translation between different SQL engines is a solved problem), but lack-of backwards-compatibility is usually the reason most good-ideas die in committee - so how come Google was successful in pushing HTTP/2 and even HTTP/3, but we're still using 1980s-level SMTP and SQL? Are all of the major RDBMS vendors so afraid of change that they're willing to forgoe winning-over millions of new customers if they can ship a usable, flexible, expressive, and safe query (and query-building) language?
I'm just blabbering at this point. Forgive me, but I need something to distract me from the outside world right now.
Interbase must have had them in 1980s since Interbase (and now Firebird) compiles ESQL queries into static BLR code. Since there's no way to compile a new query at runtime (other than by outright switching to using Interbase/Firebird's DSQL C API for a non-compile-time query), all these queries have to be parameterized to be of any use.
Their description of BLR makes it sound like an eagerly-evaluated equivalent of SQL Server’s Execution Plans, but are compiled on CREATE instead of on first-use, and are comprised of machine-native (i.e. raw x86/x64 + disk read syscalls) operators arranged together - or maybe slightly higher-level? It’s unclear how schema-binding works.
I’m curious what advantages there are to that approach anymore: databases are invariably IO-bound, not CPU bound, and things like well-maintained indexes and statistics objects are far more important when it comes to DB perf than the ISA of the query engine. I’d wager an ultramodern engine running in WASM or even interpreted Java will run faster on the same hardware than JET, I’d think.
On a related note, why are we still stuck with 4K/8K-sized pages?
Obviously they’ve been proven for tens of years now, so it’s extremely unlikely, but conceptually it isn’t different.
Parametrized Queries for a SQL library are a critical “do not ship without” feature. You do not lie and tell your user you have a safe product causing them to have a compromised system.
I hope to God you are not creating production systems anywhere.
I don't understand how anyone but the greenest of devs doesn't comprehend the importance of these kind of things. But alas, I see it everywhere, including the comment you are responding to.
I ask about SQL injection pretty regularly, and it's scary how many devs don't seem to even know what it is.
That's either a good thing: because they grew-up with database libraries designed to encourage, if not force, the use of parameterized queries, so it was never a problem for them.
...or it's a bad thing, because they grew-up in the early-days of PHP 4x, learning from PHP/MySQL tutorials written by people who'd today be considered utterly unqualified to speak at length about programming: the kinds of things that only happen when the blind were leading the blindfolded: nightmares like actively encouraging concatenating $_GET values directly into mysql_query() strings because it means having less variables, and less variables means better performance (right?!) - and they never learned otherwise.
-----
I have a pet theory that people who got started with PHP in the 2000s who are still working today are so burned from their earlier experiences that they're now the most detail-oriented and best-practices-following programmers around, regardless of the language they use today - while the people who never experienced hardship (to the extent that having to use PHP is a hardship...) become complacent, and our ever-increasing reliance on unvetted external dependencies (in all language ecosystems, imo) is going to end badly.
Or not. No idea, honestly!
Good luck building production systems without compromises.
Also, I think everyone would appreciate it if you didn’t hurl personal insults about. This aint Reddit.
Do you have examples of libraries that's actually a "parametrized query"? The python libraries I encountered (psycopg2, sqlalchemy) both seem to be "fake smoke and mirrors".
Every MySQL client needs to do something similar, due to the design of the MySQL protocol.
They are seemingly run through a stored procedure on the server side, with each parameter passed in as an argument. This has some consequences that lead to very obscure behavior, too [1].
For example, you should never create temporary tables in an SQL statement that was initialized with parameters, as they won't survive the end o the statement; they will be destroyed as soon as the innermost scope (within the stored procedure call) finishes.
I ran into this because I tried setting up a single "run this query" method in a more complicated database routine in order to keep my C# code clean... didn't work out as I'd hoped ;)
[1]: https://stackoverflow.com/a/46311328
Edit: I just realized that this still isn't a guarantee that parameters are handled as a separate structure by the underlying network protocol. I hope they are -- but I don't know how to check. Maybe the according .NET Core code is on GitHub?
Eek. "select * from classes where teacher = '?'" -> boom.
https://github.com/WordPress/WordPress/blob/4ae0744585ea9417...
They do seem to have resisted attacks for quite a while though.
That's probably why many of the libraries mentioned in this thread use "smoke and mirrors". Of course, it is quite possible to correctly escape a value by rendering it to a string first, _then_ encoding the whole thing.
To have true parameterised queries, you need to use the "binary" protocol, which many MySQL libraries don't offer support for. (MySQL also has some frustrating limitations with the binary protocol, such as not allowing a SQL string containing more than one statement to be prepared.)
Except there is no such thing as "actual parametrized queries", either as defined by the SQL standard or as supported by any of the major RDBMS vendors.
The closest thing we have are prepared statements, which are basically session-scoped stored procedures used by client-side libraries to pull off the illusion of parametrized queries.
I wouldn't be surprised that if you go back far enough, the idea of parametrized queries probably started in client libraries trying solve for SQL string soup with a dash of sprintf-like syntax sugar.
We're in the process of adding support for MSSQL in addition to the DB server we've been using for ages. This is one of the pain points. They differ in how they handle "parameterized queries" enough to be a PITA.
That would surprise me... I thought that when I send a parameterized query to PostgreSQL or MS SQL Server, most (or even all) of the query plan gets created without looking at any parameter values. And if that is true, then I think your "no such thing" claim cannot be true. If the query plan has already been created based on the unparameterized SQL string, then parameter values cannot cause it to do something crazy like drop an unrelated table.
(But I haven't read the source code to either of those RDBMS, so maybe I am about to be surprised.)
You're not sending a parameterized query. The libraries are creating prepared statements under the hood, and managing them for you. This generally requires one call to prepare the statement and fetch some sort of handle, and then executing the statement with the handle from the first call.
Part of the confusion is that it's common to see SQL libraries offer an API to parametrize queries using some sort of placeholder in the query, but it's really just a facade to perform local string interpolation along with some sort of vendor-specific sanitation scheme.
Not always. The Node `pg` module pipelines down the commands on the connection: it sends a `prepare` command, then a `bind` command, then an `execute` command. Yes, it uses prepared statements under the hood, but there is no wire round trip between prepare and bind.
Your comment peaked my interest so I looked into what postgres does, and it's pretty ingenious in my opinion. It can create either a generic plan (not looking at parameter values) or a custom plan where it looks at parameter values and then uses statistics to find the best plan.
By default, for the first 5 executions of a prepared statement uses a custom plan. If the execution costs of these different plans are pretty close to each other, from then on it just uses the generic plan (idea being that the extra cost of generating a custom plan each time isn't worth it), but if they are different enough (meaning looking at statistics can have a big speedup), then custom plans are used going forward. This behavior can be tweaked using DB flags.
In the sense that stored procedures and prepared statements are queries, sure. But your SQL statement and your values are not being sent in the same call to the database.
Is this really possible if the server considers table statistics (i.e. frequencies of certain values) when forming a plan?
They are probably missing out on many of the OTHER benefits of prepared statements. Easily 60%+ of queries can be prepared reducing overhead sometimes significantly.
From HN:
https://blog.soykaf.com/post/postgresql-elixir-troubles/
In addition to security.
connection.query("SELECT * FROM accounts WHERE username = ? AND password = ?", [{username: {username: 1}}, "secret"]);
I'm thinking they should have listened rather than getting defensive.
Also, why in the hell are values being quoted (sql parlance for escaped)? Does the mysql protocol not support actual parameterized queries?
i didn't realise SQL injection was a thing to watch out for until i'd already been working professionally for 5 years -- that maybe sounds bad, but for the first 5 years i didn't work on any projects that integrated with databases.
Don't blame that on juniors. The people writing these libraries with "magic" bullshit query parsing aren't junior. SQL injection is a very basic security issue that can be very mitigated simply with actual prepared statements.
... AND password = password = 1
This is just a cascading chain of failures.
However, proper DBMSs expose parameters as 1st class concepts through the API. That has several advantages, some of which:
1) It is more secure, as you will not have to do this dangerous escaping before invoking. Parameters are substituted by the DBMS.
2) The DBMS can better understand the "dynamic" part of a query and cache query plans.
SQLite supports all three: https://www.sqlite.org/lang_keywords.html
This is especially bad because it uses it's own escaping function and escaping in PG depends on a server configuration variable (standard_conforming_strings) that the client doesn't know about.
This behavior is barely mentioned in the docs, and the author does not really accept any suggestions or criticism.
This is where the programming community has a role to play.
When library authors blatantly brush off security issues, it's time to call out that behavior publicly and promote a secure fork. a database library should never have hidden behaviors such as theses. And "magics" such has manually building strings into a query like that s, or parsing a provided query to transform it into something else under the hood, should be turned off by default.
This is absolute madness.
https://flattsecurity.medium.com/finding-an-unseen-sql-injec...
The opening segment is the exact same piece of vulnerable code.
Still interesting.
Javascript has plenty of bad points without making without making things up.
It's not the author's fault, it's JavaScript that leads this way. Even worse is TypeScript, as its compile-time checking adds a false sense of security.
First off, most other DB drivers will use real prepared statements, so what is passed down to the DB is actually the templated query and a set of values. It looks like mysqljs actually parses and interpolates the string before sending it to the DB engine.
Perhaps more importantly, though, this bug is basically the exact type of bug that caused the log4shell fiasco: trying to be too "smart" with user provided values that should just be treated as "dumb" scalars. The fundamental way of filling out template parameters with something that can refer to DB column names is a major flaw IMO.
With prepared statements it's like saying to the DB, here's the query: SELECT * FROM users WHERE email = ? AND password = ?. And here's the data: ["whatever", "whatever"]. You can literally send whatever your heart desires for those two strings, unescaped, to the DB server and let them deal with it.
From within a language, you somehow need to restrict how you build up the query. E.g. a query builder or ORM usually makes it hard to use anything besides parameters, but they usually have escape hatches (e.g. a where method that takes arbitrary sql strings).
One trick you can do in Go is to create a type like `type sqlLiteral string`. The type is private, so other packages can't directly instantiate it, but a string literal will automatically coerce to it. Basically, you can't compute sqlLiteral from arbitrary expressions. It must be a single hardcoded string literal.
This is solely a mysqljs bug.
A library in a strongly typed language would still need to use parameterised queries or escape the strings just like a JS library. And it would be just as easy or difficult to have an incorrect sanitising function.
In fact, the specified string is not invalid input, it would be perfectly valid to store backticks in a string field in MySQL. They just need to be correctly escaped before being submitted to the database.
Again, the only bug here is the completely incorrect serialization that mysqljs does.
Sanitization is an ignorantly dangerous practice, because it obscures much safer practices.
Validation? That's for business rules. No more order items than there are things in stock, and stuff like that. If you need to do that because something in your stack needs it, then it's far past the time to reconsider your stack.
Which is exactly the equivalent of using JSON.parse(). You will find more library combinations, and for many of them the options can't be considered bugs like you do for mysqljs here, and you definitely won't be able to catch it.
The solution is parsing beyond JSON.parse(). You can't have arbitrary user provided structures floating around your program.
[1]: https://dev.mysql.com/doc/refman/8.0/en/sql-prepared-stateme...
The reason is that using a real prepared statement in MySQL (as well as most other DBMS) requires an extra round-trip, which adds latency. It also potentially adds complexity if a proxy/middleware layer is in use, since prepared statements are per-connection, at least in MySQL.
The core problem in this specific case is that this js mysql client library simply did not implement client-side interpolation correctly or securely. This is quite bad; I'm not aware of a similar problem in any other major language's most popular mysql driver in recent years.
I don't know the internals of the (binary) protocol used for communication with the MySQL server though. Couldn't one just save the extra round-trip with length-prefixed strings by sending the query together with the parameters in a single message?
[1]: http://www.gosecure.it/blog/art/483/sec/mysql_escape_string-...
at least that’s true for Postgres, I can’t speak for MySQL.
Indeed -- mysql_real_escape_string "mostly" fixes this problem by requiring a db connection as one of its args. Since the driver is usually aware of the connection state, mysql_real_escape_string can check to see if one of those exotic charsets is in-use. The issue is that there are multiple ways to change the connection charset, some of which the driver is aware of (e.g. in PHP mysqli set_charset) but some it is not (running textual statements like SET NAMES or SET CHARACTER SET).
However, generally an attacker won't have the ability to set an arbitrary exotic character set for the connection anyway... unless they already have some other sql injection mechanism, in which case it's a moot point :)
Driver documentation also typically mentions this problem. For example, here's the doc for doing client-side param interpolation in the most popular MySQL driver for Golang: https://github.com/go-sql-driver/mysql#interpolateparams (see warning in italics)
It also explicitly detects if your initial connection settings attempt to use one of those charsets along with param interpolation, and throws an error if so: https://github.com/go-sql-driver/mysql/blob/21f789cd/dsn.go#...
> Couldn't one just save the extra round-trip with length-prefixed strings by sending the query together with the parameters in a single message?
AFAIK, no, not with the traditional MySQL binary protocol. The newer "X protocol" introduced in MySQL 5.7 does allow this, but it is not widely implemented in drivers.
No, this wouldn't work, because you have to send COM_STMT_PREPARE (https://dev.mysql.com/doc/internals/en/com-stmt-prepare.html) first, which takes the SQL and returns a "statement ID". Then you can send COM_STMT_EXECUTE (https://dev.mysql.com/doc/internals/en/com-stmt-execute.html) which contains the statement ID and the parameters. Finally, you would ideally send COM_STMT_CLOSE (https://dev.mysql.com/doc/internals/en/com-stmt-close.html) to free the server-side resources for the prepared statement, although this could be "pipelined" with the EXECUTE packet.
One would expect you can do the preparation lazily as there shouldn’t even be a hypothetical performance consideration for using prepared statements.
In other words you send an execute prepared statement request, you can provide the truly prepared statement OR the templated sql string. In the latter case the server would send you the results of preparing the statement before sending you the result of execution so that you can provide it on subsequent requests.
Or use an RPC system that supports promise pipelining so that you could send the prepare request and the execute is sent with the promise of the prepare request.
With a prepared statement, there's an implication that you want to re-use it multiple times, so the DB needs to give you back some sort of handle/identifier on it. But this requires bookkeeping on both the client and server side, and that overhead is often not worthwhile on modern CPUs, especially when running fairly simple CRUD-style queries which are quick to parse. (and at least in modern versions of MySQL, the DB will track a "digest" of the query which removes params anyway, even if you used client-side param interpolation!)
So ideally DB binary protocols should offer less expensive ways of doing parameters for "fire and forget" queries, instead of having to track real prepared statements, but often they don't.
MySQL's newer X Protocol provides a way to avoid the extra round-trip, but it isn't very widely used yet AFAIK, and it still involves prepared statement bookkeeping: https://dev.mysql.com/doc/internals/en/x-protocol-use-cases-...
Not that this is a problem for a well-designed library, of course. E.g. tokio-postgres in Rust does it if you use a promise-combinator for the first poll of multiple queries inside the same transaction.
In addition to, actually using DB-level pre-compiled statements.
This just seems like bad code on bad code.
So I bumped an issue, noting this is all over HN, and offered to write a pull request for the API change proposed by the maintainers:
https://github.com/mysqljs/sqlstring/issues/60
Doug agreed to accept such a request, so I just sat down to figure out the code and a reasonable upgrade plan.
Three hours later, I could already write Doug this email (pasting it here because the issue and codebase are locked to non-contributors so I had to send it via email):
OK, I have a draft pull request ready. Of course, it's a big change and I expect to get a lot of feedback and have a few rounds of back and forth and fixups before it is accepted.
This is the plan as I envision it:
* Release SqlString 3.0.0 that has a new allowObjectValues parameter defaulting to false. This is a new major, so it shouldn't break anybody's code.
* Release mysqljs 2.19.0 (or should it be 2.18.3 for even higher adoption?) that depends on SqlString 3.0.0 but explicitly passes allowObjectValues on every call to it,with a default of true unless the user explicitly set it to false. This is a non-breaking change. This version will also add a deprecation warning whenever a ConnectionConfig is built without explicitly setting allowObjectValues, warning that the default will change in 3.0 and suggesting to set it to false unless it's needed and highlighting the need to typecheck values if it's set to true.
* Release mysqljs 3.0 that changes the default and removes the deprecation warning, so new projects get a sane default.
This involves, of course, changes to two repositories, so here they are (I can't open pull requests because I have not contributed in the past):
https://github.com/SonOfLilit/sqlstring https://github.com/SonOfLilit/mysql
(I didn't write the mysqljs3.0 patch yet to make pull request technicalities simpler, but it's trivial)
Again, I'm new to this project, am not a javascript developer in my day job, and I expect - and am prepared to handle - nontrivial amounts of feedback and requests for improvement.
For a brighter, safer future :-),
Aur