SecurityScorecard observes a 1,000-to-1 ratio of non-breached to breached companies
Dr. Bob Sohval · Cybersecurity Data at Scale // Sam Kassoumeh & Bob Sohval, SecurityScorecard (Data Driven) · Oct 26, 2017 · at 11:45
Dr. Bob Sohval, Director of Data Science at SecurityScorecard, details class imbalance challenges when training machine learning models for breach prediction.
0:00 / 0:09exact quote · 9.5s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“The number of non-breached companies, the ratio of non-breached to breached is something like, in our system or our platform, something like a thousand to one.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Dr. Bob Sohval
Experian's outdated browser usage signals poor security hygiene, says Bob Sohval
“Northwestern Mutual on the left is, is doing a pretty reasonable job of keeping their browser browser use up to date. Experian, not so much. And this is a signal that they're not paying enough attention to security hygiene in, in their company.”
Cyber breach prediction is more accurate on portfolios than individual companies
“And I would say that, ah, in general, breach prediction on portfolios is somewhat easier than for individual entities. The law of large numbers does apply here, and it's easier to make statements at a portfolio level.”
Companies with poor security scores are five times more likely to breach
“Of the probability of a breach given a poor score versus a good score, and we look at our data, and that is a ratio of five, something like that. So poor scores are predictive of breach.”
SecurityScorecard operates proprietary honeypots and sinkholes to detect corporate malware
“We are able, we operate our own bank of sinkholes and honeypots to detect malware on, on systems that are at the, on the companies that we monitor.”