Description:
INV-19060
Background
Today, commercially available IoT devices operate according to standardized and inflexible protocols. The sheer number of devices deployed in the IoT (currently over 7 Billion), will bring unprecedented levels of stress to existing IoT communications, inevitably bringing us closer to spectrum crunch. IoT devices also face serious security threats. The inflexible characteristics of existing wireless protocols are highly susceptible to hacking and jamming attacks.
In order to change the status quo, IoT systems will need to become polymorphic – in other words, they will need to (i)infer on-the-fly the physical-layer parameters currently used by the transmitter radio; and if needed, (ii) adapt their signal processing chain according to the inferred parameters.
Technology Overview
Northeastern researchers have developed PolyRF, the first deep learning based polymorphic IoT wireless platform. PolyRF is based on a novel deep learning architecture specifically tailored for the embedded RF domain, called RFNet. The platform can distinguish small-scale transitions in the I/Q domain enabling the solution of key RF inference problems. This approach can be embedded in IoT devices and applied to address practical IoT communication problems.
The team’s work is focused on taking concrete steps towards a minimalistic, protocol-free, inference-based, on-the-fly approach to IoT wireless communications. In this approach, transmitters and receivers operate using a limited set of basic rules (e.g., set of modulations) which are seamlessly changed by the transmitter at will without using control channels or headers. The receiver, in turn, infers the parameters using machine learning techniques, and then “morphs” itself into a new configuration to demodulate the data.
PolyRF takes the first step toward a vision where IoT communications are driven by artificial intelligence, and not by pre-defined protocols.
Key Benefits
- 52x latency reduction
- 8x reduction in hardware resource consumption
- Greater protection against adversarial attacks
- Implementation in off-the-shelf equipment
Commercial applications
- Commercial IoT communications
- Military communications