Original practice · fully worked
Original variant: two independent anomaly flags
One percent of data packets entering a network are corrupted. Two monitoring rules evaluate each packet. Conditional on whether a packet is corrupted, the rules act independently. Each rule flags a corrupted packet with probability 0.90 and flags an intact packet with probability 0.05. A packet triggers both rules. Calculate the probability that it is corrupted.
- A
0.0081 - B
0.1406 - C
0.1538 - D
0.7660 - E
0.9900
Answer and calculation
Conditional independence makes the two-flag likelihoods 0.90² for corrupted packets and 0.05² for intact packets. Bayes' formula gives 0.0081/(0.0081+0.002475)=0.76596, so choice D is correct.
Setup
Setup
Let C denote corruption and F the event that both rules flag the packet.
Pr(C)=0.01,Pr(Cc)=0.99 Model
Model
Conditional independence lets each two-rule likelihood be written as a product.
Pr(F∣C)=0.902=0.81 Pr(F∣Cc)=0.052=0.0025 Compute
Compute
Weight both likelihoods by their prior probabilities and normalize the corrupted-packet mass.
Pr(C∩F)=0.01(0.81)=0.0081 Pr(Cc∩F)=0.99(0.0025)=0.002475 Pr(C∣F)=0.0081+0.0024750.0081=0.7659574468 Answer
Answer
A packet triggering both rules is corrupted with probability about 0.7660.
0.7660(D)