Thursday, July 21, 2016
Police asked Dr. Jain's PRIP lab in MSU's CSE Department to create a 3-D finger of a dead man so they could access his phone because they believe the phone has clues to who murdered him. One challenge is that the fingerprint reader has a sensor to detect if the finger is live. http://fusion.net/
Wednesday, July 20, 2016
Computer vision has a long history, especially in CSE at MSU, and it is quite sophisticated, but touch will be needed to advance robotics (nice overview here: http://spectrum.ieee.org/). As an example, check out this article (http://spectrum.ieee.org/) on the recent Amazon picking challenge where a robot does the current human task of putting items on shelves or selecting from shelves to place in a bin (box). The winner from Delft does a good job, but is quite slow. A common way to pick an object is to use suction, but that doesn't work on everything so the Delft robot has a second grasping "hand." Amazon currently uses a lot of automation (check out the clever Kiva warehouse robots: https://www.youtube.com/), but they still need humans for picking. Our robotic overloads can't replace that human task. Yet.
Thursday, July 14, 2016
Current (FET) transistors have driven the computer age nearly since its inception. In spite of their death having been predicted many times, they are finally reaching their limits -- physics is a hard limit. Transistors based on carbon nanotubes have been an elusive potential substitute. This http://spectrum.ieee.org/ article gives a nice overview of the state of the art and indicates that significant progress is being made. A small 178 transistor circuit that runs has been demonstrated -- even though small that is non-trivial. One point confuses me: transistors have two properties, they are a switch and they amplify. Amplification with carbon nanotubes has been an issue and is not mentioned in the article. Is is so obviously solved that it need not be mentioned or is this optimistic article ignoring that point?
The CAN bus handles communication among computers on vehicles. It is a point of vulnerability because by law there is an external OBD connection (so backyard mechanics can work on their cars). Identifying a malicious device hooked into the CAN is critical for automotive security. A new technique has been proposed: each device on the CAN has a unique signature in its clock (called clock skew) that can be used to identify devices on the CAN. Therefore, a new, previously unknown and likely malicious device can be identified. A summary appears here: https://www.wired.com/ This latest technique joins an earlier technique that identified the regular communication patterns of devices on the CAN allowing a new one to be identified (summary https://www.wired.com/)
The FCC just approved 5G super fast cell communication. It is very different so it will require a huge infrastructure investment and time to roll it out. Simply: it is a higher frequency communication that requires "line-of-sight" which means lots (lots!) of antennas that will be "in sight" of your phone. Here is a nice article: http://www.wired.com/
Wednesday, July 13, 2016
It sounds like science fiction, but the ability to control devices directly through thought is not unreasonable. Crude interfaces have been in labs for years. The basic idea is to record brain waves when a subject thinks certain actions such as moving a finger. Recognizing those signals can be used as control. One challenge is that it is relatively easy to train for an individual, but has proven difficult to standardize. Here is an article about recent advances: http://spectrum.ieee.org/
Monday, July 11, 2016
One aspect of thinking about security is to consider side channels for attacks. Here is one where hand movements detected through a smartwatch reveal an ATM password being typed: http://www.binghamton.edu/ (Of course, the defense is simple: type with the other hand.)
Friday, July 8, 2016
There are two types of encryption: symmetric uses a fixed key (that must be shared for communication), a common usage is to encrypt your computer or phone; asymmetric constructs a key from a private key and a public key (removing the requirement to share a key) and it the basis for encryption across the Web. If you can factor large numbers into its primes (e.g. 15 = 3 * 5), you can break asymmetric encryption. Quantum computers will be able to factor the large numbers that are the basis for asymmetric encryption so encryption on the Web is endangered. Google has begun experimenting with encryption that will be resistant to an attack using quantum computers. This is important work: http://www.networkworld.com/