Billions of Intel and AMD processors compromised by attacks

Billions of Intel and AMD processors compromised by attacks
Descriptive text here
-

Researchers have identified two new types of attacks that can be exploited, compromising billions of high-end processors from Intel and AMD. The team led by computer scientists at the University of California, San Diego, detailed the findings in a paper.

Read more:

What are the attacks like?

  • An attack can target a resource called “Path History Register“, which takes care of branch destination addresses.
  • In a nutshell, in software, branches occur as programs navigate different paths. Their directions, when intercepted, provide “insights” into the data of the executed program.
  • The resource was used to anticipate future results based on history. As a result, more information is exposed.
  • The survey mentions both Intel and Advanced Micro Devices (AMD).

The researchers also successfully introduced another Specter-style attack, allowing attackers to induce branching misprediction patterns. “This manipulation leads the victim to execute unintended code paths, exposing their confidential data,” said Dean Tullsen, a computer science professor at UC San Diego.

Image: Golden Dayz/ Shutterstock

In this new study, researchers used innovative techniques to capture not only the most recent results, but also all branch results in sequential order: “We have successfully captured sequences of tens of thousands of branches in precise order“, said Hosein Yavarzadeh, Ph.D in the Department of Computer Science and Engineering at UC San Diego, the paper’s lead author.

The researchers communicated the security findings described in the paper in November 2023 to Intel and AMD. Intel has informed hardware/software vendors of the issues. Both confirmed that they will address the concerns raised in the document.

The information is from TechXplore.


The article is in Portuguese

Tags: Billions Intel AMD processors compromised attacks

-

-

NEXT What will the new Apple model look like?
-

-

-