Michael, Thanks for the support. Because we perceive this as a security problem, our intent is to be wary of any web application that relies on IIS' session tracking. The same goes for ColdFusion. I'm not too familiar with reporting bugtraq incidents, so I'm not sure where to take this thread from here. I've posted to this group, reported to CERT - if nobody else perceives this as a serious problem, I intend to let it stand as is. Thanks, Jeff ----- Original Message ----- From: "Michael J. Cannon" <mcannonat_private> To: "Jeff Jancula" <Jeffat_private>; <vuln-devat_private> Sent: Wednesday, August 29, 2001 7:44 PM Subject: Re: Web session tracking security prob. Vulnerable: IIS and ColdFusion (maybe others) > Jeff, > > Bottom line, given CERT, MS's and Allaire/Macromedia's responses (and their > history of shutting the barn door after all the cows have left), the right > answer for you is that if you and your team, after consideration, perceive > it as a security threat, then that's what it is. Due to the EULAs, and how > CERT is formed, none of these guys is vulnerable to liability. You are not > so lucky. Get your presentations in order, check and confer with your > internal and external auditors and counsel and STICK BY YOUR GUNS. Get > anyone who denies that this is a security probelem within First Union or > consultants, auditors (internal and external), vendors and management who > doubt that it is a security problem to put those beliefs on paper, explain > them and sign the paper. (For a lark , see if MS or Macromedia will send you > a written statement on letterhead that this is not a security problem). > Check with BUGTRAQ and NTBUGTRAQ. Then do as your conscience dictates. > > Your responsibility is to your shareholders and customers. Protect them, no > matter WHAT the vendors say. > > Michael J. Cannon > Ubiquicomm > "Si vis pacem, para bellum." > ----- Original Message ----- > From: "Jeff Jancula" <Jeffat_private> > To: <vuln-devat_private> > Sent: Wednesday, August 29, 2001 1:25 PM > Subject: Web session tracking security prob. Vulnerable: IIS and ColdFusion > (maybe others) > > > > SECURITY PROBLEMS WITH WEB SERVERS' SESSION TRACKING MECHANISMS. > > > > On February 20, 2001 we reported the following problem (with specifics to > IIS and SITESERVER) to the Microsoft Security Response Center. > > > > On March 22, 2001 we also reported a similar problem to Allaire (now > Macromedia) for ColdFusion. > > > > Approximately 2-3 weeks after reporting to appropriate vendors, we also > reported these vulnerabilities to CERT.ORG. > > > > PROBLEM DESCRIPTIONS: > > > > Microsoft Internet Information Server (IIS) and Site Server do not verify > that session cookie values were actually issued by the server. An Internet > user can generate their own session cookie, which will be accepted as valid > by these servers. An attacker could use cross-site scripting vulnerabilities > to generate a modified session cookie, with a predictable session value, > then use the predetermined session value to later take over (impersonate) > other users. > > > > Similarly, Allaire's ColdFusion Server does not verify that session > tracking values CFID and CFTOKEN were actually issued by the server. An > Internet user can generate their own CFID/CFTOKEN session values, which will > be accepted as valid by the server. An attacker could set CFID/CFTOKEN > values on a URL line contained in an e-mail message, or use cross-site > scripting vulnerabilities to generate CFID/CFTOKEN session cookies, with > predictable values, then use the predetermined session values to later take > over (impersonate) other users. > > > > These vulnerabilities, especially when combined with well-known cross-site > scripting vulnerabilities, could cause loss of confidentiality, failure of > non-repudiation and fraud. > > > > SUMMARY OF VENDOR RESPONSES: > > > > Microsoft agreed that we had uncovered a bug in IIS, which would be fixed > in a future release. However, they did not consider the bug to be a security > vulnerability because it requires another security vulnerability (cross-site > scripting) to work. > > > > Allaire agreed that we had unconvered a security problem with ColdFusion > and recommended that applications requiring high security implement their > own session tracking mechanisms - in other words, don't rely on ColdFusion's > session tracking (CFID/CFTOKEN). Allaire also indicated that they are > considering a redesign of their session cookie mechanism to improve > security. > > > > Like Microsoft, CERT isn't convinced that a real security problem exists. > > > > Personally, I'd like to know if the security community thinks this is a > real problem or not. > > > > BACKGROUND: > > > > When a Internet browser user visits IIS or ColdFusion hosted web sites, > the web server issues browser commands similar to: > > > > (for IIS) Set-Cookie: ASPSESSIONID=BBBBBBBBABCDEFGHIJKLMNOP > > (for CF) Set-Cookie: CFID=123 > > (for CF) Set-Cookie: CFTOKEN=4567890 > > > > The browser stores and returns the "ASPSESSIONID" or "CFID/CFTOKEN" values > with each subsequent request to the web server. IIS and ColdFusion use these > values to identify and track each user. > > > > IIS and ColdFusion do a pretty good job of generating random session > values, so that users can't guess each other's session values. However, an > attacker could force a predictable cookie value, by using JavaScript or an > <META HTTP-EQUIV> tag to override the relevant cookies: > > > > document.cookie = "ASPSESSIONID=BBBBBBBBAAAAAAAAAAAAAAAA"; > > > > or, > > > > <META HTTP-EQUIV="Set-Cookie" Content="CFID=123; path=/"> > > <META HTTP-EQUIV="Set-Cookie" Content="CFTOKEN=1111111; path=/"> > > > > Of course, the hard part is getting the JavaScript or META tag to the > victim's browser (that's were cross-site scripting comes in - a subject > covered elsewhere). > > > > ColdFusion makes this attack even easier, because it allows its session > tracking variables to be specified on the URL line. So, an attacker could > force a predictable cookie value by passing a user a link, via e-mail, > another web site, or as a bookmark. For example: > > > > http://www.MyColdFusion.net?CFID=123&CFTOKEN=1111111 > > > > Regardless of the method used, the browser will send the modified > ASPSESSIONID or CFID/CFTOKEN values for all future requests to the web > server. The problem is, the web server honors the modified session values - > as if the server actually issued them! > > > > Note that the cross-site scripting hack (using META tags or other > JavaScript commands) could change a user's cookies mid-stream, causing the > web server to detect a new session start and possibly cause the user to > re-authenticate (logon). The user would only notice that they were seemingly > "kicked out" of their session, and probably not report the incident to > support personnel. The user's previous session would eventually be abandoned > by the server; and the attacker could now intercept the new session. > > > > HOW FIRST UNION DETECTED THIS PROBLEM: > > > > This problem was detected by First Union's application security testing > team, while testing servers for potential use with First Union internal > applications. The team used various hacking tools to perform a > man-in-the-middle attack to modify ASPSESSIONID and CFID/CFTOKEN cookies. > > > > These attacks were NOT tested on live financial/production systems. > Confidential customer or employee data was not exposed during these tests. > > > > RELEVANT VERSIONS: > > > > The tested Microsoft servers were running IIS 4, SiteServer (version?) on > Windows NT 4.0, service pack 6a. Although only slightly confirmed, we > believe IIS 5 on Windows 2000 is also vulnerable. > > > > The tested Allaire servers were running ColdFusion 4.6 on Solaris. > > > > REFERENCES: > > > > "Security Best Practice: URL Session Variables and HTTP_REFERER" article > on Allaire's web site (www.allaire.com). > > > > TECHNICAL CONTACTS: > > > > Jeff Jancula, Technical Advisor, e-mail: Jeff.Janculaat_private > > Chris Howser, Technical Advisor, e-mail: Chris.Howserat_private > > Chris Hudel, Technical Advisor, e-mail: Chris.Hudelat_private > > > >
This archive was generated by hypermail 2b30 : Fri Sep 14 2001 - 10:08:16 PDT