Article Search

Main Categories

Avast Report: AI Chips Are Not Completely Secure But That’s Ok

James by James watson

Published Fri, Mar 16th 2018, 17:22 | Computers

Martin Balek, the Avast Antivirus Research Director told to media that the AI chips are not safer than the regular ones but that’s not a big deal. While both the chips are different in terms of their internal working, all contemporary artificial intelligence platforms like the NPU (Neural Processing Unit) that is available inside the Huawei’s HiSilicon Kirin 970 SoC is also a kind of specialized chip. Like the traditional silicon chip, this specialized chip is equally vulnerable to the online threats and hacking attacks, said Mr. Balek. The industry veteran further added that no platform is completely free from or secure from the hacking attacks. The statement was given on the basis of the recent issues with Meltdown security flaws and Spectre that cause Intel’s chips to illustrate that point, mentioning how both these vulnerabilities were the core outcome of the design decisions that failed to account for “all possible exploitation scenarios”. Both these vulnerabilities were neither the first nor the last example of the hacking attacks in the online world. Like the traditional chips, designing artificial intelligence platforms is also likely to cause security issues. With this, there is no chance the Neural Processing Unit and other such solutions of the upcoming time are completely secure, said Mr. Balek. He said that this is not a big problem as the cybersecurity mechanisms are always emerging together with the hackers and vice-versa. The officials from one of the leading online security services and solutions providers said that, we are not expecting AI botnets to be a realistic threat in the upcoming time. This is because of the relative shortage of such hardware that makes it least attractive to the hackers who want to target the large number of people, create technical difficulties, and cause a number of other issues. The part of the artificial intelligence in the cybersecurity is becoming huge along with the potential of the similar technologies to be attacked by the hackers. On the devices, AI computing is still limited to choose the flagship models. Such capabilities are mainly targeting to the mid-range segment in the upcoming years, with the likes of HiSilicon and Qualcomm already working on enabling them in non-premium modules. Such Android devices that have been launched this year are likely to tout AI features as their MSP (main selling points). Source: http://help-number.com/blog/avast-report-ai-chips-are-not-completely-secure/

Bio