Yee Wei Law Marimuthu Palaniswami ARC Research Network on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP) 28th March 2011, Tonkin’s 3rd Smart Grids Forum 1 ◦ 10 Australian universities ◦ Access to an extended network of over 200 researchers ◦ Australia, the USA, Europe and Asia ◦ 30+ industry linkages, e.g., ISSNIP Vision Partners: Universities National International Industry Research Institutions Funding: ARC DEST DIISR NSF DARPA 2 With UNSW Evaluate impact of largescale wind turbines penetration on the transient and voltage stability of power systems Effects of Flexible AC Transmission Systems (FACTS) devices With University of Surrey Reshaping energy demand of users by communication technology and economic incentives “Persuasive energyconscious network” usage 3 Cyber security, not to be confused with “power system security” Confidentiality Availability Integrity Outline: Recent developments Introduce new smart grid infrastructure and associated security issues What tech from sensor networks is applicable What research is needed 4 2007 2008 2009 2010 5 Organizations Guidelines/standards Nuclear Regulatory Commission (U.S.) Cyber Security Programs for Nuclear Facilities (2010) Dept. Homeland Security (U.S.) Catalog of Control Systems Security: Recommendations for Standards Developers (2011) Idaho National Laboratory (U.S.) NSTB Assessments Summary Report: Common Industrial Control System Cyber Security Weaknesses (2010) NIST (U.S.) IR 7628 “Guidelines for Smart Grid Cyber Security” (2010) NIST (U.S.) Standards review (2011) Government Accountability Office (U.S.) Report on Smart Grid Security Guidelines (2011) and more… 6 2011 CyberSecurity Watch Survey by CERT (Aug 2009 – Jul 2010) 21% threats from insiders (employees or contractors) Real-world example: Stuxnet Stuxnet was introduced via a USB stick It seeks out and spreads to machines running WinCC and PCS 7 SCADA software (Siemens) Fact 1: Insider attacks render cryptographic protection inadequate 58% threats from outsiders (e.g., hackers) 7 Fact 2: Control systems are prime targets 80% executives say their SCADA are Internet-accessible 55% say SCADA / operational control systems targeted most often 57% say security patches applied regularly 8 AGA 12: cryptographic protection of communications IEC62351: TLS encryption, security extension of DNP3, etc. Resilient control system: A system that maintains state awareness and an accepted level of operational normalcy in response to disturbances, including threats of an unexpected or malicious nature. – Rieger et al., Idaho National Laboratory Multimillion European research projects ◦ CRUTIAL ◦ VIKING 9 Part A Advanced Distribution Automation (ADA) Advanced Metering Infrastructure (AMI) Part B Wide-Area Measurement System (WAMS) 10 Part A ADA according to EPRI: Complete automation of all controllable equipment and functions in the distribution system Control center Substation RF leakage current sensor RF temperature sensor Recloser Switched capacitor bank Metal insulated semiconducting (MIS) sensor for detecting hydrogen Transmission-line robot 11 Neighborhood Area Network Smart meter Home Area Network AMI 12 Jemena, United Energy, Citipower and Powercor Interoperability Capacity Latency Interference rejection CDMA2000 GE-MDS 900MHz Open standard Proprietary 76.8 kbps (80-ms frame) 153.6 kbps (40-ms frame) 307.2 kbps (20-ms frame) Hundreds of milliseconds 19.2 kbps (80 km) 115 kbps (48 km) 1 Mbps (32 km) DSSS, 2 GHz frequency band allows frequency band re-use Transmission Nation-wide service range coverage Configuration Point-to-multipoint SP AusNet and Energy Australia Silver Spring Networks Proprietary Wi-Fi/IEEE 802.11 100 kbps 54 Mbps (802.11a) 11 Mbps (802.11b) 54 Mbps (802.11g) 72 Mbps (802.11n) Open standard Tens of milliseconds Tens of Milliseconds milliseconds FHSS, 902-928 MHz FHSS, 902-928 802.11a: ODFM, 5 GHz MHz 802.11b: DSSS, 2.4 GHz 802.11g: OFDM/DSSS, 2.4 GHz 802.11n: OFDM, 2.4/5 GHz *2.4 GHz band is crowded; 5 GHz less so 80 km Unknown 802.11a: 120 m 802.11b/g: 140 m 802.11n: 250 m Point-to-point, Point-to-point Point-to-point, pointpoint-to-multipoint to-multipoint WiMAX/IEEE 802.16 Open standard 9 Mbps Milliseconds OFDM, 3.65-3.70 GHz 20 km Point-tomultipoint * Note: ZigBee is not in here 13 Smart meter MCUs have limited computational power & memory: Vendor CPU RAM (KB) Flash (KB) Microchip PIC16Fxxxx 8-bit, 20 MHz <1 < 16 TI MSP430F471xx 16-bit, 8 MHz 4/8 56—120 Freescale V1 ColdFire 32-bit, 50 MHz 8/16 64/128/256 ADA and AMI that utilize low-capacity networks and low-cost mesh networking devices are in fact wireless sensor networks (WSNs) Notable players: founded by WSN pioneers acquired by CISCO 14 The only non-European partner in the European project “SmartSantander” To turn the Spanish city of Santander into an experimental smart city, by deploying a large-scale network of 20,000 sensors Noise mapping for the City of Melbourne To measure, monitor, understand and manage noise issues within the city The Age 15 The security of WSN is a relatively well-researched area Application Network protocol stack Network Data link Key management Physical Intrusion detection and response Many techniques are applicable, e.g., ◦ secure routing ◦ secure firmware update 16 Objective: to ensure the delivery of information Robustness: achieving objective despite hardware failures or unstable connections Resilience: achieving objective despite attacks Insider attacks: Dropping Flooding Sybil C Wormhole I’m C Wormhole B I’m B Attacker attracts traffic to itself by claiming to be 1 hop away from base station 17 RPL (IPv6 routing protocol for low-power and lossy networks) ◦ New routing protocol of the 6LowPAN protocol stack ◦ Internet Draft On top of RPL, apply “tunnel routing with support” (TRS) (codesigned with IBM Zürich) ◦ Principle: monitoring of packet forwarders, multiple paths AODV = Ad hoc On-demand Distance Vector Compared to AODV, TRS increases packet delivery rate Fraction of malicious nodes in the network Andrea Munari, Wolfgang Schott, and Yee Wei Law. Dynamic tunnel routing for reliable and resilient data forwarding in wireless sensor networks. In PIMRC 2009, pages 1178-1182. IEEE, 2009. 18 What: For updating firmware of network nodes in situ, i.e., over the air instead of via physical contact Why: ◦ Firmware needs to be updated for feature expansion or “bug fix” ◦ Sensor nodes may be inaccessible ◦ Labour cost too high Challenges: ◦ ◦ ◦ ◦ Limited computational power requires discreet use of crypto Limited data rate requires careful coordination Dynamic environment requires robustness to packet loss Insider attacks requires resilience to packet pollution 19 Limit use of digital signature verification to once Avoid flooding (broadcast storm), by exchanging advertisement, request and data messages: For robustness to packet loss, use rateless codes ◦ Instead of sending packets as is, encode them in such a way that any k number of encoded packets can be decoded For resilience to packet pollution, use Sreluge, which isolates polluters Yee Wei Law, Yu Zhang, Jiong Jin, Marimuthu Palaniswami, and Paul Havinga. “Secure Rateless Deluge: Pollution-Resistant Reprogramming and Data Dissemination for Wireless Sensor Networks,” EURASIP Journal on Wireless Communications and Networking: Special Issue on Security and Resilience for Smart Devices and Applications, vol 2011, 2010. Article ID 685219, 22 pages. 20 Part B 14 Aug 2003 North America blackout affected 50 million people Reasons include: ◦ Inadequate “situational awareness” at FirstEnergy ◦ No real-time data from Stuart-Atlanta line for Midwest ISO to work with 21 WAMS: High-capacity network of PMUs Measures voltage and current phasors (magnitude + angle) Typically 30 time-stamped samples per sec Real-time control of electromechanical oscillation, voltage, frequency, etc. Phadke and Thorp’s prototype circa 1988 Commercial products: Macrodyne’s 1690 MiCOM P847 ABB’s RES521 Aka synchrophasors, because time-synchronized using GPS 22 Energy management system (EMS) State estimator Automatic generation control (AGC) Load forecast SCADA master Other functions Wide-area measurement system (WAMS) ... PDC ... RTU/ IED Economic dispatch PMU ... ... Power network 23 Possible insider attack: inject bad data to foil detection Measurements State estimator Bad data detection Network topology processor Y. Liu et al., “False data injection attacks against state estimation in electric power grids,” Proc. 16th ACM Computer and Communications Security, 2009. 24 Attack scenario: given k compromised meters (RTUs/IEDs/ PMUs), find a vector of k false values that bypass detection Larger networks IEEE test systems 25 The attacker earns $2/MWh here The attacker loses $1/MWh here The attacker earns $1/MWh net Actually congested, faked not congested Actually congested, faked not congested IEEE 14-bus test system L. Xie, Y. Mo, and B. Sinopoli, “False data injection attacks in electricity markets,” in Proc. 1st International Conference on Smart Grid Communications, 2010. 26 A multilayered architecture with a perimeter network Firewall + VPN Stewart et al., “Synchrophasor Security Practices,” white paper It is impractical to tamper-proof a whole PMU, for maintenance reasons, etc. Even if tamper-proofing all PMUs is achievable, impractical for all RTUs and IEDs Using redundant PMUs could reduce the risk, but also costly Most (academic) research so far designed attacks under different constraints We are investigating anomaly detection methods to detect false data 27 J. C Bezdek, S. Rajasegarar, M. Moshtaghi, T. Havens, C. Leckie, and M. Palaniswami, “Anomaly detection in environmental monitoring networks,” IEEE Computational Intelligence Magazine, 2010. A. Shilton, D.T.H. Lai, and M. Palaniswami, “A Division Algebraic Framework for Multidimensional Support Vector Regression,” IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, vol. 40, no. 2, page 517-528, Apr 2010. S. Rajasegarar, J. C. Bezdek, C. Leckie, and M. Palaniswami, “Elliptical Anomalies in Wireless Sensor Networks,” ACM Transactions on Sensor Networks, vol. 6, no. 1, article 7, Dec 2009. A. Shilton, M. Palaniswami, D. Ralph, and A. C. Tsoi, “Incremental Training of Support Vector Machines,” IEEE Transactions on Neural Networks, vol. 16, no. 1, page 114-131, Jan 2005. 28 We (ISSNIP) welcome collaboration opportunities Contact: palani@unimelb.edu.au 29 Attack Prevention Containment Detection and notification Recovery and restoration 30 Synchrophasors rely on GPS GPS is vulnerable to jamming (weak signal) and spoofing Short-term solution: Enhanced Long Range Navigation (eLORAN) Long-term solution: atomic clocks A portable GPS and mobile jammer A LORAN transmitter 31 PMU PMU ... PMU PMU WAN Layer 2: Data management PDC Application Data Buffer Real-Time Monitoring Real-Time Control Layer 1: Data acquisition Layer 3: Data services Real-Time Protection Layer 4: Applications 32 Privatization of electricity market recent (‘80s) Locational marginal pricing (LMP) aka nodal pricing ◦ Case no constraint on Tx line: uniform market clearing price is the highest marginal generator cost ◦ Case congestion on Tx line: price varies with location Attack [Xie ‘10]: 1. In the day-ahead forward market, buy and sell virtual power at two different locations 𝑃1 and 𝑃2 2. Inject false data to manipulate the nodal price of the Ex Post market 3. In the Ex Post market, sell and buy virtual power at 𝑃1 and 𝑃2 respectively 4. Profit 33 Advances in sensor and networking tech driving Smart Grid Grid modernization stimulates multi-disciplinary research In progress: ◦ $100m “Smart Grid, Smart City” demo project in Newcastle ◦ Intelligent Grid: CSIRO and five universities We have the expertise in WSNs, control and machine learning, seeking collaboration opportunities 34