Cyber Attacks and Defense in Practice
An IDS examines 1 million connections a day, of which 100 are actual attacks. The IDS detects 99% of attacks and wrongly judges 0.1% of non-attack connections to be attacks. Which of the following is closest to the proportion of the alerts raised by the IDS that are actual attacks?
- AAbout 0.01%
- BAbout 9%
- CAbout 99%
Show answer and explanation
Answer: B. About 9%
Of the attacks, 100 × 0.99 = 99 are detected, and of the 999,900 non-attack connections, 999,900 × 0.001 = 999.9 (about 1,000) are wrongly alerted. There are about 1,099 alerts in total, of which 99 are attacks, so the proportion is 99 ÷ 1,099 ≈ 0.090, or about 9%. Where attacks are rare, most alerts are false positives even if the false-positive rate is small (the base-rate fallacy). Lowering the false-positive rate to 0.01% reduces the false positives to about 100 and raises the proportion to about 50%, so reducing false positives by tuning detection rules and cross-checking with other information is important for lightening the analysts' load.
- A. This is the proportion of attacks among all connections (100 ÷ 1 million = 0.01%), not the proportion among alerts.
- B. Correct. 99 ÷ (99 + 999.9) ≈ 9.0%.
- C. This is the proportion of attacks detected (the detection rate), which differs from the proportion of alerts that are attacks.
Continue: 9 more Cyber Attacks and Defense in Practice questions