The snitches sometimes nest in unsuspected places. In 2017, a group of researchers noticed that it was possible to know the destination of an Internet user just from the volume and the connection time, provided that they had previously trained a machine learning algorithm on thousands of websites. A process known as “website fingerprinting”.
Also to discover in video:
A variant of this monitoring technique was proposed in 2019 by another group of researchers. Instead of looking at web traffic, they looked at the behavior of the processor cache.
A group of researchers has now dug this furrow further. They tested five cache analysis methods on four processor architectures (Intel Core, AMD Ryzen, Apple M1, Samsung Exynos).
The result: all of these platforms are vulnerable to varying degrees. Surprise: Apple’s M1 chip is the most vulnerable of all. The researchers suspect that the heuristic mechanisms of the M1 processor cache are “Less sophisticated” than those of others, which would make the attacks more effective.
The finding is similar with Tor Browser. Some methods allow you to monitor browsing by analyzing the behavior of the cache.
In the end, the study ends on a rather dark note. Internet users cannot currently protect themselves against such surveillance. It is therefore necessary to avoid connecting to untrustworthy sites which could carry out this type of spying. The only avenue for improvement mentioned by the researchers is that of better partitioning of data, whether at the level of the processor or the operating system. But it is complex to achieve.
Source : Scientific article