Spring: Spectre Returning in the Browser with Speculative Load Queuing and Deep Stacks


METADATA ONLY
Loading...

Date

2022-05

Publication Type

Conference Paper

ETH Bibliography

yes

Citations

Altmetric
METADATA ONLY

Data

Rights / License

Abstract

There has been a substantial community effort in mitigating transient execution attacks in the web browser. Lightweight “catch-all” timer mitigations, deployed in all popular browsers, are presumed to raise the bar against these attacks. More heavyweight mitigations, such as pointer and array index masking are deployed more selectively to further make such at- tacks impractical. How secure are browsers with these mitigations taken together? In this paper, we show that a combination of new techniques allows an attacker to employ Spectre-RSB and leak sensitive information from browsers that deploy all these mitigations. First, we show that queuing up many transient loads during a single speculation window and using repeated measurements enable cache covert channels, even with jittery, millisecond precision timers. Second, we reverse engineer the newer RSB structure in Intel CPUs to find that deeper call stacks allow the attacker to hijack speculative execution for bypassing pointer and array index masking mitigations. Third, we show how an attacker can leverage memory massaging to reduce the entropy of the target secret’s memory address. Our end-to-end exploit, Spring, combines these observations to leak an access token from an unmodified version of Firefox. Our disclosure effort has led to a deployed mitigation in the latest version of the Firefox browser.

Publication status

published

External links

Editor

Book title

Journal / series

Volume

Pages / Article No.

Publisher

Workshop on Offensive Technologies

Event

16th IEEE Workshop on Offensive Technologies (WOOT 2022)

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Organisational unit

09721 - Razavi, Kaveh / Razavi, Kaveh check_circle

Notes

Funding

Related publications and datasets