A student at MIT developed an algorithm that trick AI to misidentify objects. For example, the algorithm can wrongly identify a headphone as a gun at an airport terminal. By switching a few pixels or adding noise to the image, AI can misidentify a gray tabby cat as a bowl of guacamole.
Resource: https://www.theregister.com/2017/11/06/mit_fooling_ai/
Dr. Sheryan Sen, an associate professor at Purdue University, shared his team's research paper titled, "Enabling New Interaction Modalities by Communicating Strictly Through Touch Using Electro-Quasistatic Human Body Communication."
Dr. Sen argues that wearable devices that use Bluetooth could use our body as a wire to send information securely because the signals are not getting radiated, but only transferred by touching. Using this technology you can open a computer, a home/car door, also create a second-factor authentication by just toughing with your finger. Also, it can transfer my contact information through a handshake.
As far as the applications of this technology are concerned, the possibility includes augmented reality, virtual reality, basically anything you think Bluetooth is used today around the body. This technology can replace Bluetooth technology to connect/transfer data with Human-computer Interaction utilizing low frequency Electro-Quasi Static Human Body Communication.
Currently, the team sees huge application space and multiple different large companies as well as startups have shown interest including the touch-based-ID technology and they are working with them to develop this technology and bring it to market.
Sources
https://thecyberwire.com/podcasts/research-saturday/170/notes
Joe Slowik, a Senior Security Researcher at DomainTools, discussed their research team's discovery based on their theory that identifying infrastructure in adversary operations by tracking identifiers related to major events and conflict zones can yield insights into defense and response for upcoming incidents.
During the initial investigation, the research team discovered that a document reflecting very specific themes related to the conflict in the Caucasus region between Armenia and Azerbijan. The document was masquerading as a news article and attempted to communicate with a certain domain.
The team found an unusually long string of numbers as a template object trying to attempt to communicate a domain: msofficeupate.org. Also, the template item is serving as a signifier to identify additional samples similarly constructed.
Based on the characteristics found, DomainTools researchers identified 35 domains matching the patterns associated with the initially observed malicious domain.
Overall, the adversary operations were related to political, military, and related subjects in the Caucasus region. By tracking identifiers and pivoting the investigation, the researcher were able to link to a phishing email that is state sponsored.
The lessons the team discovered is that the analysis of both the malicious documents and related network infrastructure by tracking identifiers can be used to gain insight to deploy defensive countermeasure that is coming in the near future because it is unlikely that adversaries will completely change their life cycle.
