Announcing Changes to Microsoft’s Mitigation Bypass Bounty

Today we’re announcing a change to the Mitigation Bypass Bounty that removes Control Flow Guard (CFG) from the set of in-scope mitigations. In this blog, we’ll provide additional background and explain why we’re making this change.

Mitigation Bypass Bounty Background Mitigation Bypass Bounty Background

Microsoft started the Mitigation Bypass Bounty in 2013 with the goal of helping us improve key defense-in-depth mitigation technologies by learning about bypasses. Since launching this program, we’ve awarded more than $1,000,000 in bounties and fixed numerous bypasses reported in our exploit mitigations and are looking forward to growing that number in the future.

One of the challenges we’ve faced with the Mitigation Bypass bounty program is providing clear guidance to researchers on what sorts of issues are in-scope vs. out-of-scope and what sort of cash reward can be expected. We’ve made several changes over the past few years to try to improve the situation here, such as:

1. More clearly defining payout tiers for different types of mitigation bypasses (i.e. bugs vs. design problems).
2. Being more transparent about the types of issues we are currently aware of so researchers know what types of bypasses are out of scope.

Even with these changes, we know we’re not perfect and we continue to listen to feedback and make changes to be more researcher friendly.

Impact of Exploit Mitigations on Exploitation Impact of Exploit Mitigations on Exploitation

One datapoint monitored by Microsoft is the occurrences of vulnerabilities being exploited in the wild. Microsoft has seen the amount of vulnerabilities exploited in the wild decrease steadily over the past 8 years.

We believe that part of the reason for the decline of known exploits in the wild is the increase in exploitation difficulty, which transitively affects the economics of vulnerability exploitation. We attribute a large part of the increased difficulty to Microsoft’s continued investment in exploit mitigation technologies such as CFG, Arbitrary Code Guard (ACG), Code Integrity Guard (CIG), MemGC, and so on.

Before we launched the Mitigation Bypass Bounty, we were more heavily reliant on analyzing exploits found in the wild to identify mitigation opportunities. This created lag between technique use in the wild and mitigation availability. To shorten this lag time, we launched the Mitigation Bypass Bounty to proactively learn about bypasses before they were used in the wild.

CFG has been a particularly popular Mitigation Bypass Bounty target for security researchers. Thanks to this research, we’ve learned a lot about a variety of bugs and design limitations affecting CFG. This has caused us to reevaluate the threat model that we need to defend against for more robust CFI. In order to do that, we know we will need to extend and improve the design of CFG, e.g. with finer-grained CFI, read-only memory protection, safe unwind/exception handling, and so on. We recently talked about the challenges with CFG and how our threat model has evolved (Video | Slides).

Microsoft has also received submissions and made fixes for other targets in the Mitigation Bypass Bounty, such as ACG. Researchers can expect that as we build new mitigations we will add them as bounty targets.

Changes to the Mitigation Bypass Bounty Scope Changes to the Mitigation Bypass Bounty Scope

As of today, CFG has been removed from the set of in-scope mitigations for the Mitigation Bypass Bounty. We believe we now have a good understanding of the limitations of CFG and the threat model we need to adapt the design to. We do not believe that additional research into CFG bypasses will be valuable until we’ve addressed these limitations and we would rather that researchers focus their attention on the other in-scope mitigations for the bounty. Although we are removing CFG from the bounty scope, we have no intention to remove or deprecate the feature and we still believe it is a valuable defense-in-depth mitigation. We look forward to bringing it back in scope once we’ve made improvements to CFG.

As always, we’d appreciate feedback from the community on this or any related topics.

Joe Bialek

MSRC Vulnerabilities & Mitigations Team