Search This Blog

Powered by Blogger.

Blog Archive

Labels

Showing posts with label CPU. Show all posts

New Apple Processor Vulnerabilities: FLOP and SLAP Exploit Speculative Execution

 

Security researchers have uncovered two new vulnerabilities in modern Apple processors, named FLOP and SLAP, which could allow attackers to remotely steal sensitive data through web browsers. Discovered by researchers from the Georgia Institute of Technology and Ruhr University Bochum, these flaws exploit speculative execution, a performance optimization feature in Apple’s processors, to extract private user data from browsers like Safari and Chrome.

How FLOP and SLAP Exploit Speculative Execution

Speculative execution is a technique used by modern processors to predict and execute instructions in advance, improving performance. However, flaws in its implementation have led to significant security issues in the past, such as the Spectre and Meltdown attacks. FLOP and SLAP build on these exploits, demonstrating how Apple’s latest chips can be manipulated to leak private information.

FLOP (False Load Output Prediction) affects Apple’s M3, M4, and A17 processors. These chips attempt to predict not only which memory addresses will be accessed but also the actual data values stored in memory. If a misprediction occurs, the CPU may use incorrect data in temporary computations. Attackers can exploit this by measuring cache timing differences, allowing them to extract sensitive information before the system corrects itself. Researchers demonstrated FLOP by stealing private user data, including email details from Proton Mail, Google Maps location history, and iCloud Calendar events.

SLAP (Speculative Load Address Prediction) impacts Apple’s M2 and A15 processors, along with later models. Unlike FLOP, which predicts data values, SLAP manipulates the processor’s ability to anticipate which memory address will be accessed next. By training the CPU to follow a specific pattern and then suddenly altering it, attackers can force the processor to read sensitive data. The CPU processes this information before realizing the mistake, leaving traces that hackers can analyze. Researchers used SLAP to extract Gmail inbox content, Amazon order history, and Reddit activity.

Implications and Mitigation Efforts

Both FLOP and SLAP are particularly concerning because they can be executed remotely. A victim only needs to visit a malicious website running JavaScript or WebAssembly code designed to exploit these vulnerabilities. The attack does not require malware installation or direct access to the device, making it difficult to detect or prevent.

The researchers disclosed the flaws to Apple in early 2024. While Apple has acknowledged the issues, security patches have not yet been released. Apple has stated that it does not consider the vulnerabilities an immediate risk but has not provided a timeline for fixes. In the meantime, users concerned about potential data exposure can disable JavaScript in their browsers, though this may break many websites.

These findings highlight the growing sophistication of web-based attacks and the need for stronger security measures in modern processors. As Apple works on mitigating these vulnerabilities, users should stay informed about security updates and exercise caution when browsing unfamiliar websites.

The discovery of FLOP and SLAP underscores the ongoing challenges in securing modern processors against advanced exploits. While speculative execution enhances performance, its vulnerabilities continue to pose significant risks. As cyber threats evolve, both hardware manufacturers and users must remain vigilant, adopting proactive measures to safeguard sensitive data and maintain digital security.

Are GPUs Ready for the AI Security Test?

 


As generative AI technology gains momentum, the focus on cybersecurity threats surrounding the chips and processing units driving these innovations intensifies. The crux of the issue lies in the limited number of manufacturers producing chips capable of handling the extensive data sets crucial for generative AI systems, rendering them vulnerable targets for malicious attacks.

According to recent records, Nvidia, a leading player in GPU technology, announced cybersecurity partnerships during its annual GPU technology conference. This move underscores the escalating concerns within the industry regarding the security of chips and hardware powering AI technologies.

Traditionally, cyberattacks garner attention for targeting software vulnerabilities or network flaws. However, the emergence of AI technologies presents a new dimension of threat. Graphics processing units (GPUs), integral to the functioning of AI systems, are susceptible to similar security risks as central processing units (CPUs).


Experts highlight four main categories of security threats facing GPUs:


1. Malware attacks, including "cryptojacking" schemes where hackers exploit processing power for cryptocurrency mining.

2. Side-channel attacks, exploiting data transmission and processing flaws to steal information.

3. Firmware vulnerabilities, granting unauthorised access to hardware controls.

4. Supply chain attacks, targeting GPUs to compromise end-user systems or steal data.


Moreover, the proliferation of generative AI amplifies the risk of data poisoning attacks, where hackers manipulate training data to compromise AI models.

Despite documented vulnerabilities, successful attacks on GPUs remain relatively rare. However, the stakes are high, especially considering the premium users pay for GPU access. Even a minor decrease in functionality could result in significant losses for cloud service providers and customers.

In response to these challenges, startups are innovating AI chip designs to enhance security and efficiency. For instance, d-Matrix's chip partitions data to limit access in the event of a breach, ensuring robust protection against potential intrusions.

As discussions surrounding AI security evolve, there's a growing recognition of the need to address hardware and chip vulnerabilities alongside software concerns. This shift reflects a proactive approach to safeguarding AI technologies against emerging threats.

The intersection of generative AI and GPU technology highlights the critical importance of cybersecurity in the digital age. By understanding and addressing the complexities of GPU security, stakeholders can mitigate risks and foster a safer environment for AI innovation and adoption.


Novel Downfall Bug is Targeting Intel CPUs to Steal Encryption Keys, and Data

 

Remember those severe Meltdown and Spectre CPU bugs that were discovered nearly five years ago? Intel is once again in hot water due to a severe vulnerability that impacts chips dating back years. 

The vulnerability, dubbed "Downfall," exploits a flaw in the AVX vector extensions of every Intel CPU from Skylake to the most current 12th-generation Alder Lake chips.

Macs using these processors first appeared in late 2015, with the 21.5-inch iMac, and nearly every Intel-based Mac-desktop or laptop-since then has been affected. Apple started employing its own CPUs in 2020, rather than the newer 12th and 13th-generation Intel processors (which aren't affected by the vulnerability anyhow). 

What exactly is Downfall? 

Daniel Moghimi, the researcher who identified the vulnerability, developed a microsite about it and summarises it as follows: 

Downfall attacks exploit a fundamental flaw identified in billions of current CPUs used in personal and cloud systems. This vulnerability, CVE-2022-40982, allows an individual to get access to and steal data from other users that use the same machine.

For example, a malicious app downloaded from an app store might utilise the Downfall attack to steal sensitive information such as passwords, encryption keys, and private data such as banking information, personal emails, and messages.

Similarly, in cloud computing environments, a malevolent customer might exploit the Downfall vulnerability to steal data and passwords from other customers that share the same cloud server. 

Intel was first made aware of the vulnerability last summer, but it has only now been made public to give Intel time to develop a fix. Users would receive updates from their hardware makers in the form of microcode, which Intel has only started to release for its chips to address the problem. 

Which Macs are impacted? 

It's unknown whether Macs are impacted at this time. A chip that is on Intel's list of impacted products is used inside nearly every Mac starting with the Skylake generation (starting in late 2015) that has an Intel CPU inside. Your CPU is almost probably impacted if you own an Intel-based Mac that was produced in 2016 or later (or an iMac that was released in late 2015). 

But Macs are quite distinctive. Custom motherboards and firmware have been used in Intel Macs, and some of them even had the powerful T2 processor. Until we hear from Apple, it's difficult to say for sure if any of this would necessarily stop an attack exploiting the Downfall vulnerability. 

What needs to be done next? Is a fix available? 

There isn't much you can do but wait if you own a Mac built in late 2015 or later; yet, you can be affected. If a processor microcode upgrade is required or further mitigations are required, Apple will release a macOS update. You don't need to be concerned if your Mac is Apple Silicon-based (it has an M1 or M2-based processor). 

Using only software from reputable sources is a smart idea at all times. Compared to the most recent release from a well-known company like Microsoft, Google, or something from the Mac App Store, the tool you downloaded from a website you had never heard of carries a much higher chance of virus.

'Hot Pixel' Attack Exploits Novel GPUs and SoCs to Siphon Browsing History

 

An innovative cyberattack technique known as "Hot Pixel," which targets the complex interactions between graphic processing units (GPUs), contemporary system-on-a-chip (SoC), and browser data, has been discovered through a historic partnership between the University of Michigan, Ruhr University Bochum, and Georgia Tech. 

The "Hot Pixel" attack varies from conventional security flaws, as it bypasses modern side-channel defences by taking advantage of data-dependent computation cycles in GPUs and SoCs to steal information from Chrome and Safari browsers. 

The inherent difficulties that contemporary processors have in managing power consumption and heat dissipation, especially at high execution rates, served as the foundation for the researchers' finding. This disproportion generates a distinct digital fingerprint that can be recognised and examined. 

By removing pixels from the content being displayed in the target's browser, the "Hot Pixel" attack takes advantage of these peculiarities to deduce a device's navigation history. The attackers were able to quickly determine the data being processed by observing how the processor behaved differently under various browsing circumstances.

“The rendered image of a webpage may contain private information that should be isolated from scripts running on the page,” the research paper reads. “Examples include embeddings of cross-domain content through the use of iframe elements, and the rendering of hyperlinks, which indicates whether they have been visited.”

In the Chrome and Safari web browsers, researchers ran several CPU and GPU tests. They were able to steal data based on pixels from Chrome with an accuracy range of 60% to 94%, and it took them between 8.1 and 22.4 seconds to decode each pixel. 

Sending cookies to iframe elements is prohibited by Safari's anti-pixel-stealing policy if their origin is different from the parent page of the attacker. However, the researchers found that by burying URLs to sensitive sites on their site, attackers can still exfiltrate the victim's browsing history. 

Attackers might simply ascertain whether their victim had previously visited a particular address because links are presented differently if they have been previously viewed.

The researchers suggest the following measures to stop attacks similar to Hot Pixel: 

  • Minimise devices that are thermally restricted 
  • Enforce hardware constraints by keeping systems' temperatures within acceptable ranges 
  • Remove secrets from iframes' visible content by separating cookies from cross-origin iframes
  • Get rid of unauthorised access to sensor readings (OS-level mitigation)