If you can't keep your smartphones updated, perhaps the solution to rampant security vulnerabilities is "cognitive computing technology" to block them:
Qualcomm announced that the first main application for its Zeroth neural chip will be a malware behavior analysis feature called "Qualcomm Snapdragon Smart Protect." The feature will be free for OEMs to use, but it will be up to them to enable it on shipping devices. Qualcomm's Zeroth chip uses "cognitive computing technology," which can enable "brain-inspired," on-device intelligence. The chip is meant to bring more natural interaction with devices and anticipate users' needs. Zeroth was designed to think like a biological brain and learn from its experiences in order to improve itself.
For instance, one of the first demos Qualcomm showed back in 2013 was a robot using Zeroth to find only white squares on a floor, but avoid other colored squares. The robot did this not because it was programmed in a certain specific way to reach the white squares, but because it "learned" by itself where the white squares would be. This is the main principle behind a neural processing unit (NPU) such as Zeroth, which is supposed to sit side-by-side a "traditional" CPU in devices.
The most exciting features that such a chip can provide will likely arrive later on, after developers have started using Qualcomm's Zeroth SDK to create innovative new mobile solutions that can improve people's lives. However, Qualcomm has already come up with what could be a solid use-case for Zeroth: malware behavior analysis. Qualcomm can use the brain-like cognitive power of the Zeroth platform to detect "abnormal behavior" on mobile devices, which can include zero-day malware or "transformational malware," about which anti-virus solutions either don't know or the malware was modified to bypass them (in the latter's case).
Related: Mobile World Congress 2015 Roundup
Related Stories
The Mobile World Congress is showcasing brilliant innovations in mobile technology and marketing from March 2-5.
Products announced so far at MWC 2015 include: the Samsung Galaxy S6, S6 Edge, and HTC One M9 flagship smartphones; the HTC VIVE, a Valve VR headset, and the HTC Grip; Ikea electrified furniture with wireless charging; the Sony Xperia M4 Aqua And Xperia Z4 Tablet; a 200 GB microSD, USB Type-C flash drive, and SLC/TLC Hybrid eMMC SSD from SanDisk; the Qualcomm Snapdragon 820 SoC, featuring "the Zeroth cognitive computing platform", as well as Snapdragon Sense ID fingerprint scanning; Microsoft's Lumia 640 and 640 XL mid-range smartphones, and an 84-inch touchscreen called the Microsoft Surface Hub; an HP Spectre x360 ultrabook; an Archos tablet that combines microSD with internal storage; and finally, smartwatches from Huawei and LG.
Qualcomm also demonstrated Cat 11 LTE hardware. Intel gave more details about its new 14nm "Cherry Trail" Atom SoCs, which have recently been renamed using x3, x5, x7 designations to mimic the i3, i5, i7 naming scheme of Core chips. MediaTek announced the MT8173 SoC; The Register details MediaTek's moves to compete with Qualcomm. Imagination detailed its PowerVR G6020 GPU for low power devices. Audience announced the NUE N100 multisensor processor for voice and "contextual motion" recognition. Google shared details about its plans to resell wireless service and provide Internet connectivity using balloons. Broadcom announced two new Wi-Fi combo chips for mobile devices.
For the latest coverage, check these pages at Anandtech, Tom's Hardware, The Register, and Ars Technica.
The headline that Huawei seems to want to promote is the addition of dedicated neural network silicon inside the Kirin 970, dubbed the Neural Processing Unit (NPU). The sticker performance of the NPU is rated at 1.92 TFLOPs of FP16, which for reference, is about 3x what the Kirin 960's GPU alone can do on paper (~0.6 TFLOPs FP16). Or to put this in practical terms, Huawei says that the NPU is capable of discerning 2005 images per minute from internal testing, compared to 97 images per minute without the NPU – and presumably on the CPU – using the Kirin Thundersoft software (likely a future brand name). Obviously, depending on the implementation and power use, I would expect Huawei to try and leverage the NPU as much as possible in upcoming designs.
Other details for the Kirin 970 show improvements over the Kirin 960. First is the movement to TSMC's 10nm process, from 16FF+. The Kirin 960 launched a few months before the 10nm ramp up for other high-end smartphone SoCs hit the shelves, so Huawei is matching their competitors here. The core configuration is the same as the 960, with four ARM Cortex A73 cores and four ARM Cortex A53 cores, this time clocked at 2.4 GHz and 1.8 GHz respectively. The integrated graphics is the newest Mali G72, announced alongside the A75/A55 processors earlier this year, which will be in an MP12 configuration. Frequency was not listed.
[...] Huawei's final declarations on the NPU state that it is 25x the performance of a CPU with 50x the energy efficiency, and using a new HiAI (Hi-Silicon AI) nomenclature.
I'm waiting for the smartphone that packs in a central processing unit, graphics processing unit, neural processing unit, and quantum processing unit.
Related: Snapdragon 820 SoC's Zeroth Neuromorphic Chip to Block Malware on Smartphones
Intel Announces Movidius Myriad X Vision Processing Unit
(Score: 0, Redundant) by Anonymous Coward on Monday August 31 2015, @11:44PM
Will something that only works on one vendor's SOCs be used?
If yes, seems like there is less arguement for not just using native code in Android for everything. You've already got the runtime checks, and device specific component builds.
Or, equally bad, Google abstracts all these sorts of accelerators behind the LCD and makes none of them worth using. See openCL on recent Google branded Android.
Note, I am not knocking this co-processor. I would love to play with one on an SBC running a proper GNU/Linux not Android/Linux, but the market driving Qualcomm to do this is Android.
(Score: 0, Redundant) by Anonymous Coward on Monday August 31 2015, @11:57PM
Since this needs to be enabled by each vendor, they'll probably have a say in how the "brain-inspired" malware is defined. Perhaps software from competitors that impact the vendors' in-app purchase cash flow?
(Score: 1, Insightful) by Anonymous Coward on Tuesday September 01 2015, @12:36AM
Tomorrow, 'pirated content', or 'unacceptable speech'...
No thanks.
(Score: 1, Redundant) by frojack on Tuesday September 01 2015, @01:22AM
it "learned" by itself
Machines learning by themselves is just a likely to decide that normal email should be considered harmful (because its boring and does not exercise the processor's features, and spam loaded with malware should be considered good because it gives the the processor something to do.
I doubt that you can design hardware to learn to evaluate what is going on in a computer and make proper decisions about the software load. Seems any new app downloaded would declared malignant, and killed. An the process that invited that malignancy aboard would have to go as well, so kill the app store as well.
Its hard to make things fool proof, because fools are so ingenious.
No, you are mistaken. I've always had this sig.
(Score: 3, Funny) by takyon on Tuesday September 01 2015, @02:11AM
Eventually the neuromorphic chip will just block everything including the carrier signal, after reading what Richard Stallman has to say about cell phones.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Tuesday September 01 2015, @03:26AM
fixing stuff after the horse has left the barn?
maybe install this chip on computers of "developers" that write the actual buggy code:
zeroth: "hi steve! i see you again used some un-safe fuction in your code! I am sorry i will have to chalk this up on your monthly coding review."
one wonders if any "developers" will be left after a few months ^_^
-or-
rather "let's push more crap code! we can sell more fix-me-up chips this way!"
(Score: 1, Informative) by Anonymous Coward on Tuesday September 01 2015, @07:31AM
Apparently it uses Quallcomm's own COLD neuron model described a bit here: http://www.cs.utah.edu/asplos14/files/Jeff_Gehlhaar_ASPLOS_Keynote.pdf [utah.edu]
My heretical opinion on this is that since its a spiking model, its flawed. One can never build a sentient machine out of those chips :(
However, for detecting patterns between different event sources, its great, as long as not too many sources.
(Score: 3, Informative) by mtrycz on Tuesday September 01 2015, @11:54AM
We don't even know exactly HOW the brain works, let alone replicate the way it works in silico.
Sigh.
If we only could get more of this [soylentnews.org].
In capitalist America, ads view YOU!