Signature-based AV is failing : on average across the AV industry, 40% block rates just after 0-hour of a new malware sample, rising to 60% after 5 days. sounds like the AV industry is losing, if this chart is valid. (via Terry Zink)
via:tzink malware av fail accuracy detection false-negatives scanners viruses
Sort vs. Hash Revisited: Fast Join Implementation on Modern Multi-Core CPUs [PDF] : sort-and-merge is likely to be faster on future SIMD-capable multicore CPUs RSN
sort merge hash join databases performance cpu simd multicore
#
So... the graph of detection rate over time... some of those lines sometimes descend as they move later in time. How could a sample be detected at day n, but then no longer be detected at day n+1?
eg. the brighter red line, rises to ~88%, then falls to ~86% -- how exactly is that possible? The black line (which finished at #3 position on the right) rises to ~75% before falling to ~73%. Earlier, it had risen to ~49% before falling to ~48% before then rising again. The dark blue line which is #3 for most of the time on the graph also declines after an initial hump.
Craig -- they could have some kind of heuristic based around "newness" of the sample; as the sample becomes "less new", it stops being detected. That's my guess...