Security Week 17: Rise of the Machines

Last week was rich in news on artificial intelligence. Let's start with the announcement of the publication The Information about the progress in the investigation of a fatal accident with a Uber autonomously controlled car. An accident happened in March in Arizona: the one who died in an accident crossed the road in the wrong place, and even if the driver-observer assigned to monitor the robots did not nod, he would not have had time to react. But after all, all sorts of sensors are needed: in theory, in the dark they can see better and react faster. According to The Information, the sensors actually detected the woman relatively on time, but the software made a decision not to react to the object for a while - in order to combat false alarms.

The Information note should be divided into ten: while officially in Uber they didn’t confirm or deny anything, the information came from anonymous sources. And it sounds too easy: “false positives”. It seems that the autopilot is configured so as to ignore objects such as flying bags and curb columns, otherwise it will not be an autonomous drive, but a continuous torment with constant attempts to bypass nonexistent obstacles and sudden braking. Can code optimization lead to human victims? In this case, it is not clear, but in general, probably, yes. Why not something?

At the Google I / O conference there were also a lot of announcements related to artificial intelligence. For example, a smart assistant was shown to write letters to GMail.

To be honest, it looks scary: you may not yet have decided what you will write a letter about, but Google already knows! Of course, I understand the desire of service vendors to insert machine-learning algorithms wherever possible, but maybe we can deal with letters somehow? Or not? Perhaps real progress really should be a little scary, otherwise it is not progress at all. Cars with autopilot, smartphones that can anticipate the actions and desires of the user - all this will surely make our life more convenient, but at the same time limit the possibilities of direct control over devices and services. And it will add new security risks, about which we do not yet know, and we will probably find out only by experience.

Another application of AI: the recognition of objects built-in camera.

At the Red Hat Summit conference, the Red Hat director of artificial intelligence outlined another feature of the mass development and implementation of machine-learning systems. Iron for AI will require powerful, and it's far from a fact that absolutely everything will turn out to be centralized. This may give a new impetus to the development of a whole range of hardware solutions: from processors with increased power in terms of one core to video cards and custom computing systems. In the case of the same autonomous cars, this can lead to the appearance of a very powerful computer in the trunk of your car. Although you will not be able to run Crysis on it, artificial intelligence technologies can unfold the tendency to store absolutely all of your data somewhere “in the cloud”. There will be so much data that no cloud systems will be enough. In terms of security and privacy, this is probably good news.

The most important bugfix on Google I / O.
')
One line
But at the Build conference, Microsoft told about the support of JavaScript in Microsoft Excel. Just a day after the announcement, researcher Charles Dardaman published a proof of concept on embedding miner crypovolume in Excel. That's just where he was missing.

An interesting study of the extortionist SynAck from Kaspersky Lab. Hacking computers through RDP and disguising a malicious process by accepting cloning of Windows system processes.

Disclaimer: The opinions expressed in this digest may not always coincide with the official position of Kaspersky Lab. Dear editors generally recommend to treat any opinions with healthy skepticism.