In recent years, domestic intelligent speakers have developed rapidly and have entered the stage of rapid growth.
According to the statistics of smart speaker market scale of ovicloud in the first half of 2019, the sales volume of smart speakers in China was 15.56 million units, a year-on-year increase of 233%; The sales volume was 3.01 billion yuan, a year-on-year increase of 149%.
Smart speakers have been recognized by the public for their convenience and interactivity, and potential information security problems have gradually been exposed.
In April this year, Amazon echo, Google home and apple homepod were exposed to secretly record users’ voice. In November, two researchers successfully hacked into Amazon’s latest smart speaker.
Under the background of the continuous expansion of the smart speaker market, security problems are constantly exposed. How to balance convenience and privacy?
Potential information security problems of intelligent speakers
The development history of smart speakers in China is only two years, while Amazon launched its first speaker product echo as early as 2014, equipped with Alexa speech recognition system. In 2016, Google launched Google home with assistant. Since then, Alexa and assistant have become the objects of ridicule by American talk show hosts, and artificial intelligence has become a “network celebrity”.
There is a saying in China that “people are popular for right and wrong”, which can not be used too much to describe international celebrities Alexa and assistant.
In April 2019, Bloomberg disclosed that Amazon hired a large number of employees to “monitor” the smart speaker device in real time. Even if the voice assistant of the product is not activated, the user’s private conversation will still be recorded by the device. Google has also been pointed out that there are such problems.
At present, intelligent speech products are far from being able to think and understand like human beings, and the development of speech recognition requires a large amount of early data import for the construction of “thinking database”.
The reporter learned that the speech recognition development company will manually input a large number of speech, covering a wide range of contents from content, intonation to languages.
Obviously, compared with the capacity of human brain, the “database” used in early development is far from enough. Therefore, the product side will hire employees to expand the “voice database” by collecting user conversations when users use the product to optimize the product.
As soon as the news came out, the event fermented rapidly. Major technology companies came forward one after another and said that they had improved their functions.
At present, Amazon Alexa has added a privacy option, allowing users to turn off the function of sharing voice records through this option, while Google has completely suspended the manual transcription of assistant audio.
This incident prompted the product side to improve its functions, but the “attack” from a third party never stopped the “trial” of Amazon echo.
How do smart products protect user privacy?
In 2018, Tencent blade team hacked Amazon echo in only 26 seconds.
In November this year, at the pwn2own hacker competition, two security researchers successfully invaded the latest Amazon speaker with screen – Amazon echo show 5, and won a prize of $60000.
Why are smart products that bring convenience to our lives so easy to break?
Zou Debao, senior analyst of CCID Consulting Electronic Information Industry Research Center, pointed out in an interview with China Electronics News that before using the smart speaker product, users should first enter their personal information online through the terminal. During use, the product will record the human voice and convert it into a radio wave upload platform for processing. Anyone who can conquer the platform can conquer the input voice.
As for the domestic situation, the monitoring of intelligent products in the top-level design of the overall system is not enough, and there is a lack of reasonable authoritative monitoring system for control. The security level such as setting mode and transmission route is not enough, which is a common problem faced by all product parties.
Zou Debao believes that at present, the reengineering and remodeling of intelligent terminal products need to carry out small-scale supervision and control of products through the introduction of typical cases in the early stage. For example, the blockchain mechanism is imported for multiple security encryption, and the level and number of firewalls should be high enough.
Globally, the product form of intelligent speakers is no longer limited to traditional speakers, and “screen” speakers have gradually become a new favorite in the market. The camera mounted on the speaker improves voice recognition to the level of image recognition, so how to protect user privacy from the perspective of functional design?
Take the smart screen TV product launched by Huawei as an example. The camera equipped on the screen has lifting function. If users do not need to use it, they can choose to turn it off.
In terms of hardware, Zhan Huayi, chief scientist of Changhong AI laboratory, proposed another possibility in an interview with China Electronics News, which can use sensors to try to realize some functions of the camera. The sensor can sense the outline of people and objects, but it will not reflect a clear portrait. This design can reduce the risk of user privacy disclosure to a certain extent.
“The problem is not a negative state. All emerging products must go through a long time of trial and error, and finally select the relatively best scheme,” Zou Debao told reporters.