brainsmoke

Address Space Layout Randomization (ASLR) is fundamentally broken on modern hardware due to a side-channel attack on the Memory management unit, allowing memory addresses to be leaked from JavaScript. This talk will show how.

Address space layout randomization (ASLR) has often been sold as an important first line of defense against memory corruption attacks and a building block for many modern countermeasures. Existing attacks against ASLR rely on software vulnerabilities and/or on repeated (and detectable) memory probing. In this talk, we show that neither is a hard requirement and that ASLR is fundamentally insecure on modern cache- based architectures, making ASLR and caching conflicting requirements (ASLR xor Cache, or simply AnC). To support this claim, we describe a new EVICT+TIME cache attack on the virtual address translation performed by the memory management unit (MMU) of modern processors. Our AnC attack relies on the property that the MMU's page-table walks result in caching page-table pages in the shared last-level cache (LLC). As a result, an attacker can derandomize virtual addresses of a victim's code and data by locating the cache lines that store the page-table entries used for address translation. Relying only on basic memory accesses allows AnC to be implemented in JavaScript without any specific instructions or software features. We show our JavaScript implementation can break code and heap ASLR in two major browsers running on the latest Linux operating system with 28 bits of entropy in 150 seconds. We further verify that the AnC attack is applicable to every modern architecture that we tried, including Intel, ARM and AMD. Mitigating this attack without naively disabling caches is hard, since it targets the low-level operations of the MMU. We conclude that ASLR is fundamentally flawed in sandboxed environments such as JavaScript and future defenses should not rely on randomized virtual addresses as a building block.

We are 4 security researchers who have collectively worked on 3 different attack techniques that all (ab)use memory deduplication in one way or another. There is a cross-vm data leak attack, a cross-vm data write attack, and an in-sandbox (MS Edge) Javascript data leak + full memory read/write attack based in MS Edge. In this talk we detail how memory deduplication works and the many different ways it is exploited in our attacks.

Memory deduplication is a widely applied technique to reduce memory consumption in servers, VM hosts, desktop systems and even mobile devices. Deduplication maps multiple identical copies of a physical page onto a single shared copy with copy-on-write semantics. As a result, a write to such a shared page triggers a page fault and is thus measurably slower than a write to a unshared page. Prior work has shown that an attacker able to craft pages on the target system can use this timing difference as a simple single-bit side channel to discover that certain pages exist in the system. In this talk, we show that the security implications of using memory deduplication are much more severe than initially assumed. We show that by maliciously programming memory deduplication, an attacker can build primitives to read arbitrary data from memory and even write to memory in a limited but powerful way. We exemplify these primitives using three attacks that we have recently developed. The first attack, CAIN, uses memory deduplication to brute-force ASLR’s entropy bits from a co-hosted victim VM. The second attack, Dedup Est Machina, extends CAIN in order to leak arbitrary data such as ASLR heap/code pointers and password hashes in a victim’s browser from JavaScript. Using the leaked pointers, Dedup Est Machina uses a Rowhammer exploit to own Microsoft Edge without relying on a single software vulnerability. The third attack, Flip Feng Shui, uses memory deduplication to control the placement of a co-hosted victim VM’s sensitive information on physical memory for building a sophisticated Rowhammer attack on RSA public keys. Flip Feng Shui makes cross-VM Rowhammer attacks precise, fast and reliable. As an example, Flip Feng Shui compromises the OpenSSH server of a victim VM in less than 10 minutes in 84% of the cases. We conclude memory deduplication is fatal for security in more ways than one. Speaker BIOs: Kaveh Kaveh Razavi is a security researcher at the Vrije Universiteit Amsterdam in the Netherlands. He is currently mostly interested in reliable exploitation and mitigation of hardware vulnerabilities and side-channel attacks on OS/hardware interfaces. He has previously been part of a CERT team specializing on operating system security, has worked on authentication systems of a Swiss bank, and has spent two summers in Microsoft Research building large-scale system prototypes. He holds a BSc from Sharif University of Technology, Tehran, an MSc from ETH Zurich and a PhD from Vrije Universiteit Amsterdam. Ben Ben Gras has been part of the systems security research group at the Vrije Universiteit Amsterdam since 2015. Previously, he was a scientific programmer working on the Minix operating system under Andy Tanenbaum for 10 years. Erik Erik Bosman is a PhD student in the Systems and Network Security group at the Vrije Universiteit Amsterdam in the Netherlands. He is currently working on novel side-channel attacks for leaking sensitive information from the OS and applications. He has previously developed Signal Return-Oriented Programming, a highly portable exploitation technique that abuses signal frames for creating a weird machine that the attackers can program. His minemu system is the world’s fastest dynamic taint-tracker that can be used to protect binaries against memory corruption attacks. Antonio Antonio Barresi is Co-founder and CEO of xorlab, a Swiss IT security company. Before founding xorlab, he worked at the Laboratory for Software Technology (LST) at ETH Zurich on software security related topics. His research interests are software and systems security. Before joining LST, he worked in industry as a Software Engineer, Security Consultant, and IT Risk Officer. He holds a BSc and MSc degree in Computer Science from ETH Zurich.

Other people: Ben Gras Kaveh Razavi Antonio Barresi