6thSense: A Context-aware Sensor-based Attack Detector for Smart Devices
summary
Prototype 6thSense runs on an Android smartphone, collects raw sensor streams and encodes device states as which sensors are active, trains three ML detectors and flags anomalous state sequences; evaluation uses data from 50 users, three crafted sensor-based malware scenarios and VirusTotal checks, achieves overall accuracy above 96% with Markov Chain and LMT models outperforming a ~80% naive Bayes baseline, shows low CPU/RAM/power overhead, and successfully detects trigger, leakage and data‑stealing sensor attacks.
AI-written summary · about
chapters (11)
- 0:00Intro
- 0:40Outline
- 2:09Background (2)
- 3:00New Sensor-based Threats
- 5:42Existing Sensor Management Systems (2)
- 6:41Threat Model - What Can Go Wrong
- 8:38Important Observations
- 12:466th Sense: Detection Techniques
- 13:36Markov Chain Based Detection Technique
- 16:16Performance evaluation (9)
- 22:43References
related talks
45mAnalogy Cyber Security—From 0101 to Mixed Signals
20mAttack & Defend, Defend & Monitor, Monitor & Respond
33mWhen Attackers Go Mobile-First with AI
42mIf 6,000 Mobile Malware Applications Could Talk! Ow, They Do, And A Lot!
23mA Manufacturer's Post-Shipment Approach to Fend-Off IoT Malware in Home Appliances
36mRevisiting Stealthy Sensitive Information Collection from Android Apps
11mGotta Detect 'Em All: Fake Base Station and Multi-Step Attack Detection...
24mRecent Mobile App Security Incidents from Real-World Cases