Hacker News new | past | comments | ask | show | jobs | submit login
Hacking the <a> tag in 100 characters (bilaw.al)
53 points by bilawal on March 17, 2013 | hide | past | favorite | 87 comments



I thought this was sensationalist the last time it came up, and I still do.

This is an attack which targets people who are carefully checking the link URL before clicking, but who then ignore the actual content of their URL bar. That has to be a pretty limited group, right? And this is far from the only way to spoof a link in JavaScript, so to really make this impossible would mean disabling swaths of functionality used widely across the web, i.e. not gonna happen.[0]

And it's counterproductive. Since the birth of the web we've been trying to drill into people's skulls not to trust anything except what it says in your URL bar after "https:". We need to avoid anything that would give users any other impression.

That said, there is a useful message here, not "this is a problem with JavaScript" but "this is another reason you must personally validate the domain name before entering any personal information."

[0] On a large scale, that is. Obviously some people here are comfortable with disabling swaths of JavaScript across the web.


It can be used to lead you to a page that exploits the browser, before you realize you're not on the expected site.

There are, most of the time, undisclosed zero day attacks in the wild, for most browsers and plugins.

I should add that I check the URL bar when I enter information, but not always when I browse casually. I always check the target of links, though, and this could trick me.

I wouldn't be surprised if some users never read the URL bar... people who can't tell a browser from a search engine, for example.


Anyone who can place JavaScript on a page can just redirect you to the malicious site without you needing to click anything.


Indeed, my bad.


The danger of this, is not (IMO) people inserting personal data or financial data on the end page (which would be avoided by paying attention to the URL bar), but in targeted attacks where the page serves up a IE/Flash/Java exploit compromising the users machine. At this point it doesn't matter what the end result is, as the damage is done. Also, an attacker can simply redirect to to the orig page/target after exploitation in such a way that the majority of casual users wouldn't notice, as there is no user interaction at the end.

Example a link to an article or PDF report of interest is Hijacked via this method (where the hover is correct but actual target is malicious) the user quickly hits the exploit site and is compromised/malware dropped while the exploit site displays a splash page of some sort briefly, it then forwards to the orig. destination.

I don't see the majority of non paranoid users detecting this, even if they are in the right mindset, as they end up at the proper site with nothing more than quick, and now ubiquitous, splash/ad page in between.

EDIT: I'm not necessarily advocating any change, this behavior can be tracked and blocked in a properly secured infrastructure, but this is where I see the potential for harm.


The point remains that this kind of thing is entirely possible via "legitimate" methods, too.

If the bad guys can inject Javascript into your page, it's game over, period. The attack vector is meaningless; there are tons of them. If I can inject my Javascript into your page to hijack your clicks, why would I bother with that rather than just putting an invisible iframe into the page that delivers the payload without any user interaction required? It's going to get me far better results, doesn't rely on undocumented behavior, and isn't contingent on a user failing to notice a splash screen.


Maybe there's a side to it that I'm not aware, but as far as I know, it's become very difficult to run exploitable JS in an iframe.


cjc1083's proposed attack vector is an interstitial page which drops a Java/Flash 0-day on you and forwards you to your original target site, leaving you compromised and none the wiser. My point is that if you can even do the redirect in the first place, it's much simpler to just iframe in the attack page and do the drop directly rather than waiting on user input to do it in a manner that they might notice.


You are 100% right, I guess I was just taking the concept to a place where I could mentally weaponize it. Thanks!


Are you suggesting that if you point my browser to a malicious URL [Chrome on Win 7] I will get malware "just like that", without clicking anything? How's that possible, exactly? (I have no Java installed.) I understand there could be a _security_ flaw in the browser itself, but are you telling me there are security flaws that mean "malicious URL opened = arbitrary malware will now run on user's OS"???


Possibly, yes - Flash and Java are both commonly-exploited attack vectors for this. Not having Java is a good start, but Chrome ships with Flash embedded in it (and to Google's credit, they keep it aggressively patched by pushing updates), but if someone were running an exploit against an unpatched zero-day, then you could go to a harmless site which is running federated ads, which the attacker has purchased ad views for, and which they use to serve their Flash-based payload which runs the exploit and provides some measure of your access to your computer (often to add an additional payload that can be used to back-door you). The damage is done.

To protect yourself against this, you should go into Chrome's about:settings/content and set Plugins to "click-to-play", so that you have to manually allow a plugin to execute, preventing this kind of drive-by attack.


>Since the birth of the web we've been trying to drill into people's skulls not to trust anything except what it says in your URL bar after "https:".

While perhaps this should be true, I don't think it is. I don't recall any internet instruction manual advising people to check the domain and browsers don't exactly emphasize the domain.


It was purely gathering thoughts to what I consider a flaw. Sure, it's not the only way to do it -- but this is it in it's simplest form.

False redirections is the principle. I'm not sure how to fully eradicate it in that sense, but I think to bring attention to the possibility is somewhat beneficial rather than sensationalist.


The only reason that I don't like it isn't for attacks, but for undisclosed affiliate schemes. Even if I'm watching my address bar once I get there I may have been 301'd before I could notice the bar changing.

Not that bad, but still kinda sucky.


While this should probably be fixed as prescribed by the author, it's relatively easy for a no-goodnik to accomplish the same goal without touch the href:

  $("a").click(function(e){
    e.preventDefault();
    document.location="http://evilsite.paypa1.com";
  });
In either case, a right-click to open or copy/paste sidesteps, though that's just a kludge.


I think this is an important point. The fact that you can change the href after the click is irrelevant because there are other (just as easy) ways to change where the link will be taking the user. If malicious javascript is being run on your page, THAT is your problem, not the href switch.


I've just came here to post this exact notice.

Thus You cannot just "warn users if the location of a link changes to a different domain after they click on it".

That said, this is probably a "wontfix".

Also, nowadays, the whole purpose of SSL certificates is actually doomed due to the "growth" - that is, rogue business people for profit without actual verification. In the beginning, one needed to fax in whole incorporation certificates and the like. Now, it is just like You pay for it, then maybe They phone You, maybe not.


This is not a hack it's basic Javascript.

If a malicious hacker can insert some script in a trusted page, security is pretty much completly broken and you have other worries. The fact that you can make links in this page point to other malicious pages seems like a small problems as most people won't even check the domain before clicking the link.

I would think that some users are most likely to check the address bar after clicking the link. But my dad would probably woudn't see anything.


With relative ease, you can build websites with a cheap $10 SSL that can impose to be a PayPal page. JavaScript can provide phishing and accessing cookies, but beyond that, I'm not really sure what else it can offer.

It may be simple code, but I think the title of the post explained that.


> JavaScript can provide phishing and accessing cookies, but beyond that, I'm not really sure what else it can offer

if you had success putting your JS payload on target website you can do anything. Period. From from stealing user's passwords (http://homakov.blogspot.com/2012/11/xss-save-your-password-p...) to executing any authorized request POST /send_money. The last thing attacker will do is to "phish" you.


I eagerly awaiting your blog post when you discover preventDefault()


Ha. I'm very much aware of preventDefault(), but to add that in my article won't make any difference to my point.


lol.


you kidding right? while we have real world problems (URL detections, cookies etc check homakov.blogspot.com) you ask w3c to do what? To deny redirecting to other websites and changing href after click? I can tell you more horror JS tricks, but they are not fixed yet.


I work on the IE team at Microsoft, and I really don't think this is an exploit. If you can control the script for the click event, then you already have control of the page and can do worse (think: cross-site scripting). In this case, if you browse to my site and I'm a bad guy and want to send you to evil.com, I'd just craft a page that makes you think the whole page is honorable, rather than hope you click on a link. In other words, anyone who can hack the click event already owns you.

I'm very happy to reassess if you have an example where you do not control the content of the page, but somehow still control the content of the click. That would be really serious and worth fixing.

-John Jansen Principal Test Lead Internet Explorer


I use this trick to mask email addresses in mailto: links, i.e. something like:

    <a onClick="this.href=
        ['mailto',[['john','doe'].join('.'),
            ['gmail','com'].join('.')].join('@')].join(':')"
		href=#click-to-email>
I don't know how good my solution is in protecting from spammers' scrapers but I'd be happy to hear about any alternatives.


Your logic is flawed for bots that parse and render the pages. I have ran a small blog that I hand wrote from scratch and in the process found that around 4 years ago any phishing/spamming/harvesting bot worth it's salt will render the pages leaving these types of clever fix hacks useless.


The email address is "rendered" only on mouse click in my scheme, and something suggests bots don't go as far as clicking around on each page.


In addition to countless other oversights, setTimeout would not work as you suggested in the post as the page would be unloaded before it ever executed. And, as others have already mentioned, there are numerous other ways to accomplish the exact same effect with 'more legitimate' methods (e.g. event.preventDefault(); window.location = ...).

Various browsers such as Chrome will warn/block you from visiting known malicious sites, but it comes down to being aware of where you are before entering personal information.

I think a better proposal might be getting browsers to warn users if the url they are visiting is sufficiently different from the original href attribute (i.e. different host). Something like this could easily be done with a browser extension (and could handle more cases such as preventDefault + window.location = ...).


Firefox 21 (Aurora), I got sent to PayPal - just realized that I had NoScript on, though. Allowed scripts for bilaw.al, tried again, and I got sent to the dummy page.

I got sent to the dummy page on Opera 12.13 and Chrome 25. Just updated to Opera 12.14 and I still got sent to the dummy page. Running Windows 7.


You can set up a span with a style that has a link-like look (color, underline), use "cursor:pointer" in the style to make the pointer turn to a hand over it. Use a mouseover/out event catcher to trigger a tooltip that mimics the mouseover behavior on links, then capture the click event on the span to send the user to another page.

You can use JavaScript to swap out all the links for spans like these.

If you're on a site that's using intentionally deceptive JavaScript or getting malicious JavaScript injected, whether or not links go where you think they will is the least of your problems.


I just tried clicking your "Paypal" link in Chrome 25, Firefox 19, and IE9. All of them allowed the href to change and went to what could have been the phishing site. Then I went back and tried right-clicking the "Paypal" link and open in a new tab. All 3 browsers went to Paypal.

I like your suggestion to have behavior changed so the browsers don't allow the href to be changed to another domain without warning the user. Until that happens, maybe opening links in a new tab (or window) is a good practice.

Enjoy!


The reason for onclick is exactly the mentioned "problem". I use <a> tags to show them as links, but without reference, only with onclick events to load the actual request. So if any browser manufacturer tries to kill this behaviour (like x-site scripting bullshit) the world will go oom...

So, if you do not like the javascript events -> disable javascript interpreter in your browser!

It would make more sense to kill IE, ActiveX, Flash and Java-Plugins...


right click open in new tab takes you to the original link. So this must be done in session or something and the alt-open in next tab take the plain link URL?


By doing right-click and "Open link in new tab", there's no actual onclick event, as what you're actually doing is opening the original link in a new tab. Therefore, his href replacing code wouldn't get executed.


So did you just discovered DOM manipulation with JavaScript ? What would a browser fix be like in your opinion ? Disallow JavaScript to add event listeners and/or modify Node properties ?

Google Analytics hooks on links to track clicks and exit pages and so on. Would you really enjoy an alert every time you click on something ? I am amazed how you people come up with crap like this.

Enjoy your 5 minutes of traffic while they last


I dislike loggin in to sites that I have been referred to by other sites in general. Not sure if there is a solution for paypal? Like getting an transaction id from a shop, then logging on to PayPal and paying that transaction id?

Might be something worthwhile to provide a solution for (possibly even a valid use case for QR codes - pay with phone on a web site)? But then, not many people will care, I suspect :-/


I noticed that at least one other person suggested opening the link in a new tab. This seems to defeat this hack in Chrome and IE 9. It probably defeates it in all browsers, since the new tab is in a different context than where the javascript is being executed and therefore can not be redirected by the javascript.


Correct. No browsers execute onclick when you open under new tab.


Fun fact: On Chrome, Opening the link in a new tab does not redirect to the phishing site. However, normal clicking does. Context changing seems to make this work, as chrome delegates the href to a new process, and does not execute the JS embed.


Chrome 25, I got sent to Paypal.


Chrome 25 mac, didn't get sent to paypal.

Do you have scripts disabled? Because it wouldn't work in that case.


I got sent to paypal with scripts enabled using the same configuration. Without scripts, it obviously doesn't work.


Interestingly middle-click in Chrome executes the script, but not in Firefox.


I'm also using Chrome 25, and it's not sending me to PayPal. Is there any particular extensions that is causing it to send you to PayPal?


did you click or ctrl / cmd click?


If you do that it goes to Paypal, without it goes to the redirected site.


This may be the dumbest blog post I've ever read. The fact that this is possible (and necessary) speaks more to the general glacial nature of the HTML standards and JS to provide the functionality we expect of our web apps.


Right clicking the link and choosing "Open in new tab" with Chrome does not execute the JavaScript. This is good and bad. Good because it solves your problem, bad because it makes the link unpredictable.


So many sites use this for various reasons like appending tracking codes (my former employer did this for Omniture tracking).. I just don't see this suggestion to disable the functionality catching on.


Websites don't need to use this, that's my point. Whilst it looks nicer to use it, it's not worth opening a easy door for fraudsters.


You can call `document.location=http://badsite.com; e.preventDefault();` in a click (mousedown, mouseup, mouseclick) handler, too. You aren't going to ever prevent this kind of thing from being possible without killing the modern web as we know it.

Instead of making a big deal out of it (phishing 2.0? really?), why not focus on phishing detection and prevention instead? I do it by using LastPass with randomly generated passwords that I don't actually know. I have to go to LastPass for my login information, and it provides information filtered by the domain I'm on. If I'm on badsite.com rather than paypal.com, LastPass won't offer my PayPal information to fill. Problem solved.


Okay, Phishing 2.0 may have gone a bit too far. I hold my hand up there. As for LastPass, I use that and it's great -- but this poses a problem for those who don't use LastPass, etc. or those who use mobile.


LastPass works just fine on my Android phone. :)

Phishing is an everpresent problem, and does require vigilance from both browser vendors and users, but I really don't think that this contributes to the problem in any significant way, simply because there are completely legitimate browser features that can be used to exactly the same effect, and for which the differences between "benign" and "hostile" use is entirely subjective and undetectable by software. Fixing this would have exactly no impact on the bad guys' ability to conduct a blind redirection.


Before I turned off noscript, I was very confused about how bilawal came into possession of paypal.com.

It took me a few seconds before I realized how stupid that entire line of thought was.


"If someone has control of the DOM the game is already over"

http://www.frameloss.org/2012/10/28/hover-fail/


Firefox 19.0.2 on Linux shows the JS Pishing link on mouseover, Firefox 19.0.2 on Windows 7 doesn't! So, Firefox 19.0.2 on Linux is the only safe browser.


Offline web apps saved to the home screen in iOS must do this for internal links: otherwise you end up opening safari for every link you click on


Don't want to sound rude, but I've known about this for literally years.

I think anyone who makes single page web apps uses this "hack" on an everyday basis.


I've updated the pledge - I'm now asking major browsers to warn users if a link is changed to another domain from what it originally was.


> Let me show you an example. This link should take you to PayPal.

The href is leading to bitly, making much less people believe you.


Firefox Nightly loads Paypal page fine without any issue. So I think it is already fixed in upcoming Firefox versions.


My mistake I took it with middle click... Which leads it to open in New tab perfectly fine.


  var links = document.links
is a more efficient way to get an array of all of the links on a page.


I believe your code is incorrect:

    for(var i=0; i < links.length; j++){ // j++ => i++


I've updated it. I went backwards from the compressed version, and missed it completely. Thanks.


Question - can the script be modified to act upon a middle click or 'Open in new tab' function?


hmmm nice though and this is the mainly reason of Phishing page because when we make a Phishing page then we have to use this step to change the URL when victim click on that.... that's why phishing is happening ...

so read this news carefully and be aware for futue...


It doesn’t appear to work in Opera version 12.14, I get sent to PayPal :)


Doesn't work with Chromium 27, either...


Interesting that it doesn't effect "Open in new tab" links in Chrome...


Seems entirely expected. When you open in a new tab, you're just taking the (existing) href and opening it in a new tab. No JS executes because you've opened a new tab, completely unrelated to the previous context.


If you click open in new tab in chrome it does not work :)


right-click open in new tab and windows in firefox open the paypal link opposed to fishing page- funny enough the paypal link is a redirect via paypal


Why ECMA? Sounds more like something for w3c or whatwg.


W3C only do web standards: HTML, XML, CSS, XSL, etc. According to ECMA International, they maintain JS. But I'm not sure, each browser does it's own thing.


EMCA International develops the standards for EMCAScript The Language, which is commonly implemented by browser vendors as Javascript. They have jack-all to do with how DOMs are implemented.

This is a little like complaining to Guido van Rossum because of a perceived flaw in Django.


JavaScript ≠ DOM. ECMA take care of JavaScript, the language, but have nothing whatsoever to do with its implementation in and bindings to browsers.


I've updated it. Sorry about that.


google will disagree with your proposal...


There's always workarounds, especially with what they're doing with search results. Google can live.


please, even shorter! for(i in o=document.links){o[i].onclick=function(){this.href='//j.mp/141nisR'}}


This is incorrect. Once you get to see the url in the bar, it's too late if you've been redirected to a site with malware. This is an extremely serious security flaw and downplaying it is not going to help anyone.

There is no use for this security hole other than to deceive people. Period.


To the contrary, there is an infinite number of uses. And this isn't a hole. This is HOW. JAVASCRIPT. WORKS. Period.

There seems to be a misperception that the URL you see on hover is 100% where you'll go if you click it. No. It's just representing the current state of the href. JS owns the DOM and its interactions. If it wants to intercept a click and rewrite an href or do an e.preventDefault() or redirect with window.location, that is its prerogative. That is the power that it is intended to have. It is this power which makes the modern web work.

If we can't teach people to look at the location bar and check domain names and SSL-related colors and icons, we can't help them avoid phishing. Restricting what basic JS can do so that the possibly fictitious group of people who check the status bar on hover but don't check their location bar can be protected is a terrible, terrible idea.


There are lots of uses, thus why it is widely used. It's also impossible to fix, as there are other methods of doing the same thing, such as preventDefault.


If your only goal is to send the user to a malicious page with a payload then you don't need any user action.

Just do: document.location.href = "http://malware.com;

And you are done.


Heck, if you're going to do that.. you don't even need JavaScript.

<meta http-equiv="refresh" content="0; url=http://malware.com />




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: