Microsoft update secretly fixed two 'severe' bugs
DNS spoofing patch released on the sly
Updated A recent security patch from Microsoft silently fixed two severe bugs that were never disclosed even though they posed a risk to many of its customers, a security researcher said.
MS10-024 fixed two flaws that made it possible for adversaries to intercept victims' email messages sent by Exchange and Windows SMTP service, Nicolás Economou, a researcher with Core Security said. But the bugs - which made it "trivial" to spoof responses to domain name system queries - weren't disclosed and were never assigned a Common Vulnerabilities and Exposure identifier, sparking criticism that the critical bugs weren't properly disclosed.
Instead, the Microsoft bulletin referred only to a denial of service vulnerability that was rated either "important" or "moderate." By underplaying the risk of the threats being fixed, MS10-024 didn't give IT admins adequate information in deciding when, or if, to install the patch, Core said.
"These vulnerabilities were fixed by the patches referenced in MS10-024 but were not disclosed in the vendor's security bulletin and did not have a unique vulnerability identifier assigned to them," the Core advisory stated. "As a result, the guidance and the assessment of risk derived from reading the vendor's security bulletin may overlook or misrepresent actual threat scenarios."
Microsoft issued the following statement:
"The purpose of security bulletins is to help customers accurately assess their risk as part of their planning. We do not include comprehensive information about all variants addressed as part of our investigation, but the information we do provide around severity, and risk accurately pertains to the vulnerabilities discussed in the bulletin and any variants that are addressed as part of the investigation. In other words, no variant represents a greater severity than the vulnerability discussed in the bulletin."
Core described the undisclosed vulnerabilities as "two severe bugs" residing in both Microsoft Exchange and the SMTP services included in the 2000, XP, 2003, and 2008 versions of Windows. They made it "trivial" for attackers to pull off DNS cache–poisoning attacks first described in the early 1990s and made famous two years ago by researcher Dan Kaminsky.
One of the bugs causes vulnerable versions of Exchange and Windows to generate DNS queries using incremental transaction ID numbers. That made it easy for malicious DNS servers to guess the values and send fraudulent responses. A second flaw failed to verify that the value of the ID field in a DNS response matched the corresponding DNS query packet previously sent.
Machines that installed the patch, which was released last month, have been purged of both bugs, but users would have no way of knowing that from the bulletin that accompanied it. Indeed, the only hint of a fix comes in a FAQ section that said: "This update also includes a defense-in-depth change for Microsoft Exchange 2007 and Microsoft Exchange 2010 that adds additional source port entropy to DNS transactions initiated by the SMTP service."
Core criticized that disclosure, saying source port entropy isn't the same thing as the value of the transaction ID field used in outbound DNS queries. It also said verification of ID responses is mandated by section 9.1 of RFC5452.
"Core does not consider the two bugs reported to be 'security-in-depth' fixes and points out that there is an amount of literature to support that opinion starting with Core's first published security advisory on DNS query ID prediction and ending with Dan Kaminsky's over-publicized DNS poisoning technique which in 2008 Microsoft considered bonafide bugs that required public disclosure using their own CVEs as disclosed in MS08-037," it said.
More information and commentary from the Breaking Code blog is here. ®
This posting was updated to include comment from Microsoft.
Perhaps more likely:
... maybe it's just that it has so many bugs that they were trying to fix one bug and fixed two others by accident.
security through obscurity
is no security at all
"potential zero-day attacks based upon such disclosure"
If it's been disclosed, any attack based on the disclosure is, by definition NOT zero-day.
Of course, whenever a patch is released, the bad guys can try to exploit the hole where ever people are slow to patch. The point of the descriptions and severity rating is to allow administrators to make their own rational decisions about how quickly they should patch. Microsoft lied to their customers in saying MS10-024 fixed "moderate or important", but not "critical" holes, and therefore undermined the ability of those customers to make the best decision.
They also made a strong incentive for the bad guys to carefully reverse-engineer future patches to find other more-important-than-advertised holes.