PHP - my CMS security - php

I have my own cms, I'm exploring the security options which I should implement and study.
All my CMS pages are checking of user permissions based on the UserID stored in Session. I was thinking that how much important is it for me to "secure" the forms inside this pages? I mean the users have the opportunity to write SQL statements so if they would want to like use SQL injections it is not really that hard...
So my thinking is to have only the logon page secured + protection against session hijacking, is that correct or am I missing something?
Forms on the front page, query strings etc are also secured.
Thank you for any inside on this.

Securing web apps is a never-ending task. People find vectors to exploit every day (quite literally) through a combination of tools and methods. When building a CMS, you've got to be particularly careful with:
Users being able to log in as someone who they are not
Users being able to remotely inject permanently into a page (especially when combined with above)
Users being able to modify or extract contents of the database they are not supposed to gain access to
All these have to do with form submission, or more generally, userdata entry. It boils down to one simple precept: never ever trust an user.
The first one you have to worry about, and by far the most dangerous, is SQL injection. It's what a lot of companies fall victim to every year, and it is quite in-depth, though most attacks can be prevented by using parametrized queries (TRUE parametrized queries. PDO by default emulates them, there's a config flag to turn on true parametrization). This will effectively do most of the SQL-relevant sanitization for you, leaving the "sensible sanitization" to you (i.e. making sure that a date is actually a date). This protects you from what people tend to call first-order SQL injection - in other words, direct entry.
Second-order is more difficult. It involves storing something in your database that will then get executed in another statement. The CISCO guide has a good idea of how stuff happens.
That's all for SQL to defeat the large majority of script-kiddies.
In terms of content-filtering, you want to use a whitelisting approach rather than a blacklisting one. Remove everything you do not trust or cannot recognize. In terms of HTML, there's a wonderful tool called HTMLPurify that does exactly this.
Also
Don't make the mistake my boss did. Check for the session immediately, not just before rendering the page after the controller has run.

Related

Potential dangers of using unprepared SQL queries when not processing user input?

Everyone knows or should know parameterized queries help to protect against SQL injection. All of the tutorials and documentation I have seen have revolved around using prepared SQL queries to process form input. But what about when there isn't any form input? I.e. a subsequent query after a user has been logged in such as $stmt = "SELECT theme_preference FROM users WHERE user_id = '1234'";
$query = mysqli_query($conn, $stmt);
Is there any possible way an attacker could exploit this? (Let's say I'm using PHP).
The question is not whether the source of the data written in a SQL query is a http form. It's not even if it's from the current request.
The question is whether you trust the source of the data. And that may be a complex question.
You obviously do not trust something that comes from the current request. You also don't trust something that may have come from an earlier request, like for examples fields in a database that are modified by request data. But you also may or may not trust other fields in your database. For example you have IT ops staff, or DB admins. Do you trust them to not inject some kind of an XSS or secondary SQLi attack into a database field to steal user credit card data, which is stored in an audited table, so they cannot just go in and dump it without being noticed? If they injected javascript or a clever SQLi in the right place in a table that is not audited, they may steal credit card info by exploiting the vulnerable application, then change it back and remove all traces.
Also an application may have different sources for data, other systems may for example upload files (say XML) on APIs, data from those will be processed, some of it will eventually make it to the UI or used in SQL queries. If you trust those sources, you may choose to not implement protection against SQLi or XSS. But why would you, when it is easy? Multiple layers of defenses is always better than walking on thin ice.
So in short, the question is trust. If you absolutely trust the source of the data (like for example because it's static, hard-coded, or for some other reason), that's fine to use it directly in queries. But in case of SQL injection, using it properly (ie. in parameters) is so easy that you should just do that.
Also consider future changes. You are writing it in a SQL string without parameters because you know that it's safe now. Months pass, somebody adds a new feature, modifies your query, adds a few more parameters, one is from the request. But the pattern was already there, he will probably just copy-paste and go with the pattern - and your application is vulnerable.
My final point is static security scanners, those that look at your source code. Pretty much all of those will flag your code for SQLi if a variable is included in the query string itself without using parameters. That may of course be a false positive, but I doubt you want to bother with those findings, when you can avoid them in the first place.
So sometimes it's not just about the technical exploitability, there are other aspects too.

Why don't we have a specific database only for authentication and another one for the rest?

I wanted to know why isn't easier and more secure to have one database service, dedicated only to user authentication (withholding usernames and passwords) which would be well secured (inputs well sanitized etc.)
And another database service for all the other data in the website which would be possibly less secured and available for serving web content.
Wouldn't it be a good way to prevent a lot of the SQL injections happening these days?
No. There is a difference between sanitizing inputs and protecting passwords. SQL injection can happen in any query, and allows someone not only to steal, but also to destroy data. This should be prevented at all times.
Regarding usernames and passwords, there are additional measures to take. For one, don't store passwords at all, but always store hashed versions, hashed with a salt. Furthermore, you might put the authentication in a different database or even some authentication service if you like, but like I said, that's extra.
Of course, if you got a big bunch of data that is read-only from a website perspective, you might store that in a different database, or at least make the website connect using a database user that doesn't have any rights to modify the data. That way, the data is better protected against destruction, because errors on the website, SQL injection and even leaking of the username and password will not lead to destruction of the data. But this also goes for both the user credentials as the 'less important' data.
You seem to miss two points :
the rest of your data also should be secure (or else there's no point in securing the application)
you don't store passwords. Never. You only store a hash built from the password, the username (optionally) and a record specific salt. So it's not so important to secure that base, it only is done to prevent impersonation
As there's no reason to secure in a greater manner that part of the DB, just ensure the whole is correctly secured, it's so much easier to manage one DB than two TB. Trying to keep different types of data in different RDBMS and to maintain their coherency is a nightmare.
Lastly, note that this shouldn't change anything regarding to mysql injections. Just don't allow those injections by using prepared statements.
This is a tough quesiton. Indeed you can have two databases one for sensible data, the other for "furnishing".
Let's put it this way: even if you use a database only for "furnishing", what happens if that database get hacked ? Will your users have to see "THIS SITE WAS HACKED" in the section or...? :)
EDIT: What is protection? Protection is when you buy a coffee and you don't want your friend or someone to drink it. Then what if you buy a coffee and a muffin. Would you protect only one ? Would you protect both ? Which one is more important for you ? If you're like me and most people you would say both. What you made me understand by commening is that muffin (users) is more important than something else (furnishing).
Why would you need protection, if you don't care about the coffee ? Probably because you don't really value coffee enough to protect it. Then let me say this: the money you used to buy your coffee were worth when they were still money, but now they're coffee, what's changed apart from their shape ? Nothing.
Now let's get back to our case: if you have a simple database which contains only "furnishing" texts like "TITLE", "DESCRIPTION" etc. Having them lost isn't bad ? Won't you have to go there and fix them ? Won't you have to WORK to get everything pretty ?
Not to speak that most vulnerabilities related to a database might get pretty worse than just losing only ONE database. Having a sql injection issue might get all database destroyed, that's the danger you're trying to avoid. So, it's almost pointless to try to secure a database more than another. Try to secure everything instead, you'll be sure that everywhere you go, nothing could happen.
So you fear a special inmate, and you secure him A LOT, and trascurate that old cell. You look away, the inmate in the old cell is now out of his cell! And worse, he freed the special inmate (and everyone else). That's the kind of security hole a sql injection imply.
I think you need to look deeper into your RDBMS protection layers to get a better picture why they are better in the same instance.
Let's take for example your MySQL database:
an instance is made out of 3 + databases (mysql, information_schema, and all others you create).
so you already have a separate database repo that is handling the auth + user + pass - and that is your MySQL database.
There are ways to develop applications that are hacker free - when working with MySQL most of the coders use open SQL code (inline SQL) - and this is the problem - hackers will take advantage of this! To fix this use stored procedures (encapsulated).
Inside database privileges are a problem as well - lazy developers or stupid DBA will give extra rights for application users. Your idea is redundant from many points of view. More instances will cost you more money, more resources, etc.

mysql injection damages?

I Just noticed that my mysql_real_escape_string function is not inside a '' in some of my php scripts and it was vulnerable to injections and things like sleep(30) executed on my production site.
I am going the PDO route and implementing the prepared statements after lots of reading here. but this is not implemented yet.
Few questions, I see in my logs that lots of injections where done by people online but I can not see any damages. the user that the site runs to do sql queries has update/select/delete/insert only privileges.
But I am woried things like sleep(30) and what not works and if they did any damages I am not seeing?
Can you tell me where to check for damages or was I safe for at least major damages?
Can they have changed hidden mysql settings or system settings?
By the way, I tried to run latest updates on centos 6+ linux and php.
Thanks
edit:
just to clarify, the database is empty almost and i am not worried about the data being there and the passwords are hashed sh512. so the data inside is not important since this is a new application i am writing. but i am worried if they changed anything on the system or the db i should be worried about. some of the injections i see have java etc but the log is huge and its going to take time to go over it. i also see some schema strings in the injections.
now the question is can they have read my schema info or modified them? why does functions like sleep are working if it is a restricted user? what other functions could they have run?
note i have other DBs in the same MySQL. should i be woried about those?
by '' i mean:
select * from dbname where id=scaped_string
i should have put it in quotes
Checking for damage done to your data is dependent on the kind of data you have in your database. If after careful inspection you don't see anything wrong, then there is probably nothing wrong. If your data is of any decent size, this will be difficult or impossible.
There are many automated bots roaming the internet looking for code vulnerable to SQL injection attacks. Their attempts are probably what you are seeing in your logs. Just because an attempt was made does not necessarily mean an intrusion occurred.
Also keep in mind that you won't necessarily have evidence of data being stolen. The best way to determine this would be to take your server logs and replay them on a copy of your current server, checking to see if you get any data back.
Without any further information, we have to assume the worst case: An attacker is able to read, write, and delete arbitrary data in your database, and possibly even files on your file system, which may lead to compromise of your whole server (e.g. command execution via PHP file, privilege escalation, etc.).
Now let’s have a closer look at the possible impact:
Since PHP’s MySQL extension does not allow multiple statements, so called stacked statement attacks are not possible. So the remaining possibilities depend on the actual statement verb and the context, i.e. the clause inside the statement:
SELECT statement:
Reading any accessible data (e.g. sub-queries, unions, boolean-based blind, etc.)
Reading files (LOAD_FILE())
Writing files (… INTO OUTFILE …)
UPDATE statement:
Obviously updating any accessible data, possibly not just in the current table
Reading any accessible data (e.g. sub-queries, boolean-based blind)
DELETE statement:
Obviously deleting any accessible data, possibly not just from the in the current table
Reading any accessible data (e.g. sub-queries, boolean-based blind)
INSERT statement:
Obviously inserting arbitrary data
Reading any accessible data (e.g. sub-queries)
Some of these may not have a direct impact but may be used to exploit other vulnerabilities. In the end, it depends on the actual vulnerable statement.
As far as I know, there is not hidden mysql setting other than sanitizing your inputs to protect your data from being injected.
You said, you saw lots of injections done by people into the site, and yet you can not see any damage, maybe they were just injecting a plain HTML tags, or trying XXS attacks, because the real mysql injection, can delete your tables, update, reset every data subjected to mysql injection, so if I were you, I would implement PDO right away.
If you have anything like admin panel you should check for webshells and other backdoor-like tools on your server, because attacker could easily read your credentials from appropriate table. And, ofcourse, look for any changes in your pages (look for iframes and suspicious JS code).
Worst case scenario is executing INTO OUTFILE in writeable directory and then accesing it via local include or directly.
But, first of all, before worrying you should consider this as most common automated sql-injection checkers (bots you might say) and if you don't see any damage - mose probably there was no intrusion. But be careful - most intruders nowadays don't look for any visible damage, most probably they will inject some malicious code in your pages (like iframes with their exploits).
So, don't be too paranoid, but still cautious :)

Is there anything I need to worry about besides SQL injections and XSS attacks?

I'm finishing up my first "real" PHP application and I am trying to make sure it is secure. I'm kind of afraid that since I'm not an "expert" PHP programmer that I might be missing something huge, so I would like to give you some information about my application and hopefully you can tell me whether or not that is the case. So here we go:
I'm using a CMS to handle user authentication, so I don't have to
worry about that.
After discovering PDO shortly after starting work
on my application, I ported all of my code over to using prepared
statements with PDO.
I am escaping all form and database data (even stuff I think is safe) which is being output with htmlentities().
My application does use a session variable and cookie variable, but the function of both is very unimportant.
I have designed my form processing functions in such a way that it doesn't matter if the form were somehow altered, or submitted from off-server (i.e. I always check the data submitted to ensure it's valid).
I have done my best to make all error messages and exception messages polite but very obscure.
I'm forcing pages with sensitive information (such as the login page) to be served over https.
When I first starting writing my application, I didn't know about prepared statements, which is kind of a huge deal. Have I missed anything else?
OWASP maintains a list of the Top 10 Most Critical Web Application Security Risks (warning, PDF download). This is from 2010, but I think it still applies, perhaps even moreso now.
Injection and XSS are the top two, but you should certainly be aware of the other 8. If you are using an existing CMS, many of these may already be considered, but the more popular the CMS the more you risk running into vulnerabilities because of black hats trying to find holes in it.
If you are not storing critical data like credit cards, order history, addresses, and even emails, then I wouldn't worry too much about your site being affected as long as you are taking the basic precautionary measures (and it sounds like you are).
If you are concerned about security issues, a good resource is the OWASP - Top 10 Application Security Risks
The most important thing to take care in web applications(specially PHPs') is Data Validation of all the inputs taken from the user which are further saved in your database.
For a secure application, all the transactions should be done on HTTPS. For a secure cookie management Secure and HTTPOnly cookie should be implemented.
Some more points I don't see mentioned yet. Most of these are not related to code - I am not sure if you only wished for things related to code, but I'll mention them anyway.
Backups (user data). should be self-evident
Version control. If you have a big bug, you want to have access to the previous version.
Audit trail, alarms and logging. If you do get into trouble, how will you find out? Are you able to track down what happened? if you know something is wrong but don't fully know what, are you able to diagnoze the issue?
Hosting. Where are you hosting? Do you have adequade bandwidth and monitoring? What happens if you get DOSed? Are you able to block out unwanted traffic?
Caching. Can you change it if needed?
There's always one thing left. Availability :) There are three aspects of security:
Confidentiality (Noone can read what they don't have access to)
Integrity (Noone can change any data what they should have to and you have to be able to detect if it happened even so)
Availability (The data, application whatever has to be available)
You pretty much did a nice job and took care of the first two (credentials, prepared statements, htmlentities...) but none of them will help against a DoS attack. You should be able to detect if someone slaps your site and ban the attackers ip from your server. Although this can be done in PHP (still much better to kick the attacker at the first line of php than let them initialize the framework, database connections etv.) it can be done mre effectively in lower layers (for example: nginx/apache, iptables, snort).
However what you're asking for that usually comes to the matter of risk management. In a real application you're not able to be prepared for all the possible attacks, edge cases etc. What you need to do is classify all the risks by the probability and the impact ( http://www.jiscinfonet.ac.uk/InfoKits/infokit-related-files/Resources/Images/risk-matrix ). With this you can focus on the most important (highest) risks first and probably you can completely ignore the lower bottom part.
SQL Injection and XSS are the most prominent Hacking methods.
You are covered from SQL Injections if you use prepared statements.
Also, if htmlentities() on everywhere you display HTML you should be safe.

Is using parameterized in just the login script enough?

More of a risk assessment question than technical.
So i have been reading a lot about protecting against SQL Injection. Lets assume i have a large web application that is mostly unprotected and i need to make some improvements to protect against this problem. Read large as lots of SQL interaction using dynamic queries but on few (less than 100 registered users).
My question is... If i have already secured the main login script (the only publicly available user input) using parameterized queries, is it really that necessary to do the same work for the rest of the site? I mean if a potential attacker can't log in, how much other damage can he do?
Assuming of course none of my registered users have malicious intent.
Assuming of course none of my registered users have malicious intent.
Never assume that :) If your users have different rights, one with almost no right could find the way to get admin access... And so on.
Also, protecting against SQL injection does not make your login bulletproof and someone could bruteforce through it or simply have sniffed packets from your trusted users. Make sure you protect your whole application, do not assume that the entry point you planned (login script) is the only one available...
And please use SSL if your website contains or requires any sensitive data from your users.
What makes you sure that none of the legitimate users won't attempt to attack your system?
I'm glad my bank don't think that. Some-one would have emptied my account by now, just by using sql injection attacks.
And even then you should be careful about accidents. If you use parameterised queries, someone using a ' in a value won't cause problems.
It allows type safety in your parameters.
It engenders better programming habits.
It allows execution plan caching. (Without it, every query with a different parameter will look like a completely different query. Not only preventing caching of plans for those queries, but over-filling the cache so plans that really can be re-used get discarded.)
Not only is it more secure, it's more reliable. Other than experimentation and throw-away code, I can think of no project where I would have condoned not using parameterised queries.
Never trust your users!
Even if you do (and nobody can sign up either), there's always the potential of CSRF for instance; an attacker would lead users of that system to a seemingly harmless page with an image. That image would have the following code:
<img src="http://yourserver.com/profile?user_id='; DROP TABLE users" />
My hacking skills are practically zilch, but you can imagine what would happen :)

Categories