1. Ranveer Chandra (Microsoft Research)
Low Cost and Secure Smart Meter Communications using the TV White Spaces
Omid Fatemieh (UIUC)
Ranveer Chandra (Microsoft Research)
Carl A. Gunter (UIUC)

2. Advanced Meter Infrastructure (AMI)
AMI: integral part of smart grid
Reconfigurable nature and communication capabilities of advanced (smart) meters allow for deploying a rich set of applications
Automated meter reading
Outage management
Demand response
Electricity theft detection
Support for distributed power generation

3. Existing AMI Communications
ISM bands
Crowded in urban areas
Short distances not suitable for rural areas
Cellular links
Expensive and low bandwidth
Crowded in urban areas and limited in rural areas
Proprietary mesh network technology reduces inter-operability and impedes meter diversity
Idea: Use white spaces for AMI communications
Propose a secure architecture that yields benefits in cost, bandwidth, and deployment
Design architecture such that high bandwidth and long transmission ranges yield benefits in cost, bandwidth, and deployment

4. White Spaces
White spaces are unused portions of TV spectrum (54-698 MHz)
Excellent long-range communication and penetration properties
FCC’s recent rulings (Nov ‘08, Sep ‘10) allows for unlicensed communication in white spaces
Spectrum sensing helps with identifying and assessing quality of unused channels
Standards and research prototypes
IEEE Wireless Broadband Regional Area Network
Point to multipoint architecture
Typical range: km (but up to 100 km)
WhiteFi [BahlCMMW09 - Sigcomm ‘09]
Wi-Fi like connectivity over white spaces for up to 2km
Adaptively operates in most efficient chunk of available spectrum
Both require centrally aggregating spectrum sensing data
Several scenarios in white space networks require the aggregation of spectrum sensing data. For example:
-In order to form a network over the white spaces, the CRs need to periodically report sensing results to a base station. The base station is in charge of collecting the readings from the CRs and determining the areas of primary presence. This centralized approach has been endorsed by the IEEE WRAN standard draft [3], CogNeA [1] and recent research publications [9].
- Collaborative sensing refers to the process of combining spectrum sensing results from cognitive radios for the purpose of primary detection. The main benefit of this approach is the mitigation of multi-path fading and shadowing effects, which improves the detection accuracy in highly shadowed environments [20]. In addition, it allows for relaxation of sensitivity requirements at individual CRs [45].
- Crowdsourcing of spectrum reports from white space devices can be used to build a nationwide database of spectrum availability. Such a database can be used to augment the white space geolocation database mandated by the FCC [2] or to learn spectrum usage as part of the recently passed Spectrum Inventory Bill [4].

5. Proposed Architecture
Utility operates WhiteFi networks
Utility buys service from independent service provider
Large number and geographical spread of meters -> great for spectrum sensing -> utility can offer data to provider

6. Benefits High data rates (at low cost)
Single hop from meters to WhiteFi base station
No need for complex meshes
Saves energy used in mesh maintenance and routing
Large base of sensors for the provider
Lowers cost for service provider
Lowers service cost for utility
Lowers cost for providing broadband service to rural areas
Facilitates additional meters deployments in rural areas (particularly along power lines)
Saves energy used in mesh maintenance and routing

7. Challenges and Security Considerations
Cost of equipping meters with CRs and antennas
Will be lowered with large-scale production
May be lowered for utility by contract with provider
Limited availability of white-spaces
Unlikely in rural and suburban areas
Can use ISM or narrow licensed bands as backup
Primary emulation / unauthorized spectrum usage attacks
Transmitter localization [ChenPR – JSAC ‘08], Anomaly detection [LiuCTG09 - Infocom ‘09], Signal authentication [LiuND10 - Oakland ‘10]
Malicious false spectrum sensing report attacks
Vandalism: falsely declare a frequency as free
Exploitation: falsely declare a frequency as occupied

8. Detecting False Reports
Particularly important for AMI
Errors will disrupt AMI communication
provider cannot only rely on meters
Meters owned by a different entity (utility)
Meters may not be well-distributed, or get compromised
Must use additional sources for spectrum sensing: mobile units, consumer premise equipment, or deployed sensors
Sensors have unknown integrity and or get compromised
Detecting false reports
Based on propagation models [FatemiehCG – DySPAN ‘10]
Based on propagation data [FatemiehFCG – NDSS ‘11]
O. Fatemieh, A. Farhadi, R. Chandra, C. A. Gunter, Using Classification to Protect the Integrity of Spectrum Measurements in White Space Networks, NDSS 2011.
Summary: A key enabling technology for forming networks over the TV white-spaces is the aggregation of spectrum availability data from multiple sources. However, this aggregation is vulnerable to maliciously misreported measurements. We propose a technique that uses trusted propagation data to build an SVM classifier, which is subsequently used to detect violations. Our work eliminates the need for arbitrary assumptions about propagation models and parameters. Evaluations using FCC and NASA data show our technique is effective.

9. Data-based (Classification-based) Detection
Model-based schemes: not clear which signal propagation models, parameters, or outlier thresholds should be used
Idea: Let data speak for itself
Provide natural and un-natural signal propagation patterns to train a machine learning SVM classifier
Subsequently use classifier to detect unnatural propagation patterns -> attacker-dominated cells
FatemiehFCG – NDSS 2011

10. Evaluation Transmitter data from FCC Terrain data from NASA
Flat East-Central Illinois
Hilly Southwest Pennsylvania (Stress Test)
Transmitter data from FCC
Terrain data from NASA
House density data from US Census Bureau
FatemiehFCG – NDSS 2011

11. Pennsylvania Stress Test Results
20km by 20km area
Data from 37 transmitters in 150km radius
Train using data from 29
Test on the data from 8
Represent un-accounted fading and other signal variations: add Gaussian variations with mean 0 and std. dev up to 6 (dB-Spread) only to test data
FatemiehFCG – NDSS 2011

12. Summary AMI communications a key part of smart grid
Proposed communication architecture that offers improvements in bandwidth, deployment, and cost
Discussed security and reliability challenges
Identified exploitation/vandalism as important attacks and proposed techniques to detect them
References
O. Fatemieh, R. Chandra, C. A. Gunter, Low Cost and Secure Smart Meter Communications using the TV White Spaces, ISRCS ’10.
O. Fatemieh, R. Chandra, C. A. Gunter, Secure Collaborative Sensing for Crowdsoucing Spectrum Data in White Space Networks, DySPAN ’10.
O. Fatemieh, A. Farhadi, R. Chandra, C. A. Gunter, Using Classification to Protect the Integrity of Spectrum Measurements in White Space Networks, NDSS ’11.

13. Backup

14. Standards and Research Prototypes for White-Space Communications
IEEE standard draft
Wireless broadband regional area networks over TV white spaces
Point to multipoint architecture (base station to up to 255 clients), with the possibility of having repeaters in between
Each access point covers km (typical) but up to 100 km
Antennas 10m above the ground, similar to TV antennas
Support for co-existence between cells
WhiteFi [BahlCMMW09 - Sicgomm ‘09]
Wi-Fi like connectivity over white spaces for up to 2km
Adaptively operates in most efficient contiguous chunk of available spectrum
Client to access point communication: using modified stock Wi-Fi cards
Requires a separate antenna and board for spectrum sensing
For spectrum allocation, both techniques support spectrum sensing and using transmitter databases