Enhancing Autonomous IoT Device Pairing with Different Sensor Types
This research explores enabling autonomous IoT device pairing using various sensor types to streamline communication between smart devices. It addresses the challenges posed by heterogeneous sensor types and the need for efficient, secure, and human-independent pairing solutions. The study focuses on verifying co-presence through mutual sensor events, introducing fingerprints, and establishing confidence measurements to differentiate between legitimate and attacking devices.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
E N D
Presentation Transcript
Do You Feel What I Hear? Enabling Autonomous IoT Device Pairing using Different Sensor Types Remy Pham
Outline 1. Background 2. Motivation 3. Perceptio Implementation 4. Evaluation 5. Conclusions
Background Goal of IoT - communication between smart devices Ideally autonomous Modern IoT topologies contain diverse sensor types Heterogeneity in sensor types can lead to complicated communication between different devices
Background Intercommunication between sensors varies in difficulty Same sensor types speak the same language Different sensor types cannot Devices should pair confidently
Motivation IoT communications contain private, sensitive data Modern incarnations of IoT require human intervention Increasing number of devices in typical IoT topology
Motivation Existing pairing solutions Human-in-the-Loop-based pairing Seeing-is-believing 2D bar code authentication Bluetooth Secure Simple pairing Wi-Fi Protected Setup Requires excessive human intervention over time
Motivation Context-based pairing Extracts entropy data from surroundings No need for human involvement Current issues Existing solutions don t handle heterogeneity Compatible sensors still possess discrepancies in data
Motivation How do we get different sensors to communicate? How do we allow communication that is efficient? How do we deal with attackers? (co-presence)
Perceptio Verify co-presence by considering mutual events experienced by different sensors Introduce fingerprints Confidence measurements that distinguish between legitimate devices (LD s) and attacking devices (M s) Become stronger over time
Perceptio Protocol Overview 1. Generate fingerprints from devices raw event data 2. Treat differences in device fingerprints with fuzzy commitment scheme 3. Generate a shared session key
Perceptio Protocol Overview
Perceptio Assumptions/Constraints Physical boundaries of house provide a trust boundary Attenuation caused by interior walls is managed
Perceptio Generating Fingerprints Data type used to compare observed events Need commonalities between device data Individual event timestamps are not sufficient Event cluster time intervals are!
Perceptio Generating Fingerprints Trade-offs Removes some signal content Benefits Tolerates hardware differences Tolerates signal attenuation Tolerates heterogeneity No need for tight time sync across devices!
Perceptio Fingerprint Entropy Uses entropy of event timings in environment to bootstrap trust Entropic threshold defined by
Perceptio Fuzzy Commitment Scheme 1. Initialization 2. Key Agreement 3. Key Confirmation 4. Confidence Score Check
Perceptio Fuzzy Commitment Scheme - Initialization 1. Device A broadcasts its ID 2. Device B receives the message, wants to pair with A a. Sends RQST_TO_PAIR to A (along with B s ID) 3. A sends a RSP_TO_PAIR message to B 4. They both enter the Key Agreement Phase
Perceptio Key Agreement Phase - Fingerprint Extraction Algorithm Determine event cluster Create fingerprints Devices compare fingerprints
Perceptio Key Agreement Phase - Determine Event Cluster 1. Signal Detection a. Pre-processing b. Thresholding and Signal Detection 2. Event Clustering a. Feature Extraction b. K-Means Clustering and Elbow Method
Perceptio Key Agreement Phase - Signal Detection Pre-processing Compute moving average for signal smoothing, noise reduction
Perceptio Key Agreement Phase - Signal Detection Thresholding and Signal Detection Set bounds for the signal Use signal lumping to group parts of event signal into a single event
Perceptio Key Agreement Phase - Event Clustering Feature Extraction Separate perceived events via clustering Distinguish using common time-domain features Maximum amplitude, duration, etc. Choose set depending on event visibility Geophone, microphone, accelerometer: amplitude, length
Perceptio Key Agreement Phase - Event Clustering K-Means Clustering and Elbow Method Takes as input k cluster groups Outputs data points to similar clusters Use Euclidean distances between data points Minimize distances No need for training!
Perceptio Key Agreement Phase - Create fingerprints Concatenate bit-representations of intervals
Perceptio Key Agreement Phase Devices compare fingerprints Pairing devices A and B perform pairwise search for matching fingerprints Compute set of commitments as a function of fingerprints The secret set is to be decoded by another device B Needs to be close enough to fingerprint of A
Perceptio Key Confirmation Phase Devices create a shared session key from kABfrom k Increase respective confidence Every time Key Agreement Phase and Key Confirmation Phase are completed Verify the confidence score of each device is greater than the confidence threshold
Perceptio Confidence Score Check Phase Each device decides the other is contextually verified As long as overall confidence score is above a threshold
Perceptio Vulnerabilities to consider: Shamming attack Eavesdropping attack
Perceptio Eavesdropping Attack - attacker equipped with 1. Normal level resources a. Standard, off-the-shelf IoT sensors 2. Medium level resources a. Sensors more powerful than normal 3. Powerful level resources a. Asymmetric capabilities
Evaluation Set of LD s inside environment Set of M s outside attempting Shamming attack
Evaluation Average Fingerprint Similarity - Legitimate Devices
Evaluation Average Fingerprint Similarity - Attacker Devices
Conclusions Legitimate devices yield 94.9% average fingerprint similarity Attacker devices yield 68.9% average fingerprint similarity
