IoT Technical Report #2
By Romain Picard on Monday 3 April 2017, 10:00 - Permalink
Welcome to the second edition of the IoT Technical Report, a monthly selection of technology news on the IoT world. You will find here the most relevant discoveries and publications in the following domains : Machine learning, distributed systems, energy, embedded systems, and hardware. You will also find this report on my twitter account.
How the CIA Can Hack Your Phone, PC, and TV (Says WikiLeaks)
THE NSA, IT seems, isn’t the only American spy agency hacking the world. Judging by a new, nearly 9,000-page trove of secrets from WikiLeaks, the CIA has developed its own surprisingly wide array of intrusion tools, too. On Tuesday morning, WikiLeaks released what it’s calling Vault 7.
Amazon AWS S3 outage is breaking things for a lot of websites and apps
Amazon’s S3 web-based storage service is experiencing widespread issues, leading to service that’s either partially or fully broken on websites, apps and devices upon which it relies. The AWS offering provides hosting for images for a lot of sites, and also hosts entire websites, and app backends including Nest.
Multiple vulnerabilities found in Wireless IP Camera (P2P) WIFICAM cameras
TL;DR: by analysing the security of a camera, I found a pre-auth RCE as root against 1250 camera models. Shodan lists 185 000 vulnerable cameras. The "Cloud" protocol establishes clear-text UDP tunnels (in order to bypass NAT and firewalls) between an attacker and cameras by using only the serial number of the targeted camera. Then, the attacker can automaticaly bruteforce the credentials of cameras.
ARM DynamIQ: Technology for the next era of compute
ARM currently has a portfolio of CPUs that our partners have deployed to drive the innovation in compute over the last two decades – together we have changed the way the world thinks about compute, a world where over 3.5 billion people rely on ARM-based devices for their primary compute access. However, the challenges we just highlighted give us an opportunity to redefine compute on the CPU again, changing how and where computing happens to meet future market demands. Let me show you how ARM engineers are innovating for the future…
Mythic launches a chip to enable computer vision and voice control on any device
Hardware that responds to voice commands is already out there and probably in your hand or house right now. Whether it’s a smartphone, smart speaker or wearable, it has to connect to the cloud to deliver answers. Now, a startup called Mythic (formerly known as Isocline) is launching a chip and software that will change all that, putting voice control, computer vision and other kinds of AI into our devices locally, no cloud required.
Laying a trap for self-driving cars
We spend a lot of time and words on what autonomous cars can do, but sometimes it’s a more interesting question to ask what they can’t do. The limitations of a technology are at least as important as its capabilities. That’s what this little bit of performance art tells me, anyway.
Embedded Linux Conference
The Embedded Linux Conference (ELC) has been the premier vendor-neutral technical conference for the past 12 years for companies and developers using Linux in embedded products.
After a decade of collaboration, the conference extended its scope in 2016 to include user-space developers, the people building applications on embedded Linux, and will be the preeminent space for product vendors and kernel and systems developers to collaborate with these influential technologists.
Evolution Strategies as a Scalable Alternative to Reinforcement Learning
We’ve discovered that evolution strategies (ES), an optimization technique that’s been known for decades, rivals the performance of standard reinforcement learning (RL) techniques on modern RL benchmarks (e.g. Atari/MuJoCo), while overcoming many of RL’s inconveniences.
Learning to Remember Rare Events
Despite recent advances, memory-augmented deep neural networks are still limited when it comes to life-long and one-shot learning, especially in remembering rare events. We present a large-scale life-long memory module for use in deep learning. The module exploits fast nearest-neighbor algorithms for efficiency and thus scales to large memory sizes. Except for the nearest-neighbor query, the module is fully differentiable and trained end-to-end with no extra supervision. It operates in a life-long manner, i.e., without the need to reset it during training.