Conference Presentation
Kevin Fu
- Future wireless adoption in medical devices is anticipated to be driven more by marketing considerations than by specific clinical safety needs.
- Current and near-term malware threats to medical devices are projected to remain low-level and automated, potentially relying on malware versions from 10 to 15 years ago.
- Emerging risks for Internet of Things systems are expected to stem from devices making automated decisions based on unverified sensor inputs.
- Specific threats include electromagnetic interference or intentional radio frequency signals that could cause devices like defibrillators to misinterpret sensor data.
- Research indicates that arbitrary waveforms may be transmitted to nearby devices by manipulating a computer's memory bus to act as a transmitter.
- Acoustic tones delivered at specific resonant frequencies, such as 8 kilohertz, could cause drones to lose control by confusing internal MEMS accelerometers.
- The memory row hammer vulnerability may be exploited by adversaries to gain supervisor privileges by inducing corrupted bits in memory.
- Future security architectures for sensor-driven systems will likely require mechanisms to verify that sensed data corresponds to actual physical phenomena.
- A machine learning-based product designed to detect anomalous activity and malware in ruggedized hospital environments is currently under development for deployment on difficult-to-secure systems.
- The team plans to continue expanding and hiring to develop products targeting high-consequence devices and hospital environments.