Qualcomm explained how its Smart Protect technology works in a blog post yesterday. Unlike the traditional malware identification techniques, such as checking file signatures against a known database of hostile apps, Smart Protect is based around advanced machine-learning behavioural analysis baked into the processor of the phone.
The actions taken by running apps are analysed in real time as instructions run through the chip. Any suspicious activity can trigger an instant alert and classification of severity. Smart Protect can report back to the operating system to inform the user of why the behaviour was deemed potentially malicious.
Qualcomm writes that the technology has been designed “from the ground up” for mobile devices, powered by the company’s “uniquely deep access to the entire hardware and software stack” of its Snapdragon 820 processor. The system is based around Qualcomm’s Zeroth machine learning platform and its heterogeneous computing abilities, where different types of processor core are used on the same chip to gain increased performance across a variety of different tasks.
Because Smart Protect will be a component of the physical processor, it can continue to operate even if the device’s operating system is successfully attacked and compromised. Qualcomm also cites its always-on status as a key advantage; unlike cloud-based anti-virus services, Smart Protect will be operational from when the phone is turned on until it is turned off, even without an Internet connection.
Qualcomm says that it is working with noted mobile security providers Avast, AVG and Lookout to provide interfaces to Smart Protect in updates to their mobile products. This will allow users to see when Smart Protect detected suspicious behaviour from within their usual mobile security app.
The technology will begin to be seen from early next year and will initially appear in new smartphones using the Snapdragon 820 processor. Qualcomm has not said whether it intends to put Smart Protect onto its lower-end chips in the future as it currently relies on the Zeroth machine learning of its flagship product for 2016.