Data-Driven False Data Injection Attacks Against Power Grids: A Random Matrix Approach
read more
Citations
Moving-Target Defense Against Cyber-Physical Attacks in Power Grids via Game Theory
Hierarchical Blockchain Design for Distributed Control and Energy Trading Within Microgrids
Moving-Target Defense Against Cyber-Physical Attacks in Power Grids via Game Theory
Analysis of IoT-Based Load Altering Attacks Against Power Grids Using the Theory of Second-Order Dynamical Systems
Analysis of IoT-Based Load Altering Attacks Against Power Grids Using the Theory of Second-Order Dynamical Systems
References
Power Generation, Operation, and Control
Probability and Measure
MATPOWER: Steady-State Operations, Planning, and Analysis Tools for Power Systems Research and Education
Two decades of array signal processing research: the parametric approach
Related Papers (5)
A Novel Sparse False Data Injection Attack Method in Smart Grids with Incomplete Power Network Information
Frequently Asked Questions (16)
Q2. What are the future works mentioned in the paper "Data-driven false data injection attacks against power grids: a random matrix approach" ?
In the future, the authors will explore how the results of this work can be used to address the defense problem against these attackers ( e. g., MTD ).
Q3. What is the objective function of the FDI attack vector?
The objective function of (17) gives the number of non-zero elements in the FDI attack vector while restricting the attack vector to a m−dimensional subspace of the estimated column space, where m ≤ s ((6)).
Q4. How many independent trials are used to calculate the eigen modes?
The FDI attacks are constructed using the estimated eigen modes and their detection probability is computed by averaging the BDD’s detection results over 1000 independent trials.
Q5. How long would it take to obtain a decorrelated measurement?
Under an4optimistic assumption of obtaining a temporally decorrelated measurement every minute, the attacker would require a measurement time window of 4900 minutes, or approximately, 80 hours, for the ratio of M/T = 10.
Q6. What is the simplest way to construct an undetectable FDI attack?
In data-driven FDI attack, the attacker strives to construct an undetectable FDI attack by learning the system parameters using the accessed measurement data.
Q7. What can be done to achieve the attack’s detection probability?
In particular, the vector cs can be tuned by the attacker to achieve his objectives, such as minimizing the attack’s detection probability or causing a desired attack impact.
Q8. What is the proof of Lemma 1?
In particular, the result of Lemma 1 states thatasymptotically, the estimated eigen vectors ûi are orthogonal to uj , j 6= i, since Ωs is diagonal.
Q9. What is the ML technique used to recover the estimate of the system state?
The estimate of the system state, denoted by θ̂[t], is recovered from the measurement vector z[t] using a maximumlikelihood (ML) technique [12]: θ̂[t] = ( HTH )−1 HTz[t].
Q10. What is the detection probability of a FDI attack constructed using?
for each estimated eigen mode i, the FDI attack is constructed as a = ciûi, where ci is set to ci =√ τω̂i/µ̂i ,such that is satisfies the constraint of (15).
Q11. What is the limiting eigen value distribution of z?
To construct a data-driven FDI that can bypass the BDD with a high probability, the attacker must first estimate the number of eigen values/vectors, s, that can be reliably recovered from the measurements {z[t]
Q12. What is the deterministic mapping between eigen values of z?
the result [14] states that for all µi > √ p, when M,T →∞,M/T = p > 0, there exists a deterministic and one-to-one mapping between eigen value of the sample covariance matrix (Σ̂z), i.e., between λ̂i and µi.
Q13. How can the attacker enhance the BDD-bypass probability?
The authors showed that in this regime, the attacker can enhance the BDD-bypass probability by constraining the attack vector to a lower-dimensional subspace spanned by the accurately estimated basis vectors.
Q14. What is the attack vector by a[t]?
Denote the attack vector by a[t] ∈ RM , the sensor measurements under attack by za[t], where za[t] = z[t] + a[t], and the BDD residual under attack by ra[t] = ||za[t]
Q15. What is the limitation of the RMT spiked model?
despite this limitation, the authors will show by simulations in Section VI that RMT spiked model results are accurate for various power grid bus configurations as long as the number of sensor measurements M is large compared to N.
Q16. How can the attacker solve the FDI attack?
Using this, the data-driven FDI attack can be formulated as the following optimization problem:min cscTs (I− Ω̂s)cs (12)s.t. ||∆θ||22 ≥ τIn the optimization problem (12), the attacker designs cs to minimize the probability of detection among all attacks that satisfy ||∆θ||22 ≥ τ.