Presentation by Alberto Costa (20 April 2021) introducing RE-OPT, a robust-optimization framework for defending systems against malicious attacks using a trilevel model (defender — attacker — user). The talk formulates the defender’s problem as maxx minz max_y p(y), discusses computational challenges, proposes RE-OPT (maximize attacker budget Γ subject to guaranteeing a performance threshold π), explains solving via duality and bisection over Γ, and illustrates with a max-flow example and computational experiments (small/medium/large networks). Practical outputs include an example fortification plan and performance comparisons against benchmarks.

Questions this document answers

Alberto Costa — 20 April 2021.

10 M$ (10 million dollars).

π = 6 (flow units).

x*: 4.2 M$ to line (s,3) and 5.8 M$ to line (3,t) (total 10 M$).

Γ* ≈ 8.7 M$ (tolerance 0.1 M$).

More from Alberto

PDF

cv_academic.pdf

3mo ago
PDF

redart_content.pdf

2y ago
PDF

66090ff78f86b.pdf

2y ago
PDF

ISCO2010_Costa.pdf

2y ago
PDF

TOGO2010_Costa.pdf

2y ago
PDF

WIMS2011_Costa_2.pdf

2y ago
PDF

WIMS2011_Costa.pdf

2y ago
PDF

Bsc_Costa.pdf

2y ago
PDF

FRS-DSTA-Technical-Report-final_v2.pdf

2y ago
PDF

Strategy_to_Validate_Sensor-Placement_Methodologies_in_the_Context_of_Sparse_Measurement_in_Complex_.pdf

2y ago
PDF

1-s2.0-S1572528617300634-main.pdf

2y ago
See all content