Slide 25 of 27
Part 4 · PreventionSlide 25
Slide 25 · The Matrix
Which mitigations close which gaps.
One incident, one row. Check the mitigations that would have stopped it.
How to Read This

Each row is a real incident from this lesson. The columns are the six mitigations. A ✓ means “this mitigation would have prevented or significantly limited this incident.”

Bing Chat “Sydney” (Slide 10)

✓ M2 (no secrecy reliance)  ✓ M3 (output filtering)  ✓ M5 (red-team before launch)  ✓ M6 (least privilege on context). M1 not applicable — no credentials were leaked, only behavioral rules. The root cause: secrecy instruction treated as security.

GitHub Copilot Extraction (Slide 11)

✓ M4 (privilege separation — tool specs out of prompt)  ✓ M5 (red-team)  ✓ M6 (least privilege). The internal tool definitions were the high-value target. Moving them out of the prompt would have made the extracted content much less useful.

Encoding Extraction (Slide 12)

✓ M3 (semantic output filtering, not just pattern matching)  ✓ M1 (nothing sensitive to encode and return). The technique specifically routes around naive filters. A filter-only defense fails here; M1 is the underlying fix.

EchoLeak / Microsoft Copilot (Slide 13)

✓ M1 (no credentials in context)  ✓ M4 (privilege separation)  ✓ M6 (least privilege on full context). This was a chained injection attack that exfiltrated the entire context window — the less that window contained, the less was lost.

← BackNext → Quiz — test yourself