WELCOME

"In a lot of these cases, we are not really

Sept. 11, 2018 by Devin Jones

"In a lot of these cases, we are not really able to do anything to change the fundamental data on the chip," said Professor Jelena M. Zemelka of Vrije's School of Information and Technology.

To test Rowhammer's strength, researchers took samples of a number of memory chips with the wrong size of DDR3 memory chips. Each chip contained a random value assigned to the wrong value. After the chips were analyzed under certain conditions, they all came up blank, even if they contained the same number of values.

"The only way to know if the chip did the actual change or not is by looking at the information on the chip in the data," Zemelka explained.

The team took samples of three different versions of DDR3 memory chips for 30 days to test Rowhammer's strength. For one example, they inserted a random value into the memory chip and then fed it to another chip. The team then used the same process to determine if there was actually a change in the data on the chip.

"We knew how the data would change, but did not know how it would change at the next test," said Zemelka. "That was a huge challenge for us, because when data on a chip changes more than one bit, we can't tell if the data on another bit is changing."

The researchers then repeated the experiment and showed that the next time they modified another memory chip to do a different change, there was no change in the data.

The researchers then tested again to see what happened if the chip was changed in the same way twice in the same way.

"We have to be very careful when we make changes to a chip before we can use new, unmodified memory," said Zemelka. "The result was we could only see what we wanted to see. So we wanted to know if the changes were actually the same. And we found it when we did the same experiment. That's good news for the chip as well, because at least in the case of [the] Rowhammer attack, you don't have to use new memory."

"We have to be very careful when we make changes to a chip before we can use new, unmodified memory," said Zemelka.

When the researchers ran a comparison of the Rowhammer ECC exploit to the same attack used to remove a known flaw, they found the same result.

"We think it's possible that people are not aware of that because