Friday, April 25, 2008

Why You Can't Shoot the Same Foul Shot Twice

In the late '80s and early '90s, before the Cleveland Cavaliers became Lebron's squad, the team's fortunes rested heavily on the back of the steadfast and reliable point guard Mark Price. One of Price's claims to fame: his run at Calvin Murphy's streak of 78 consecutive free-throws-made.

In April 1993 Price missed what would have been his record-tying shot from the charity stripe; later that month, Michael Williams of the Minnesota Timberwolves would end up breaking the record in a game against the Utah Jazz. Most scientists, who over the past decade may have hypothesized why the all-time NBA career free-thrower (he made 90.4 percent of his shots) missed that crucial attempt, would likely have blamed the inconsistency of so-called on-line movements--the neural activity and muscle contractions that occur after Price bent his knees and started his motion. A new study in this week's issue of Neuron, however, reports that another factor came into play: the brain does not plan the execution of a shot in exactly the same fashion each step of the way.

"The punch line with [our] paper," says Stanford University electrical engineer and neuroscientist Krishna Shenoy, "is that this is the first evidence that neural activity--brain activity that happens well before the movement ever begins--has a lot to say about the variability or the exact movement that you're going to get."

Shenoy and his team studied two rhesus monkeys as they made a simple, practiced movement--reaching to touch a target--to determine whether so-called "off-line" activity had any effect on the variability of each movement. First, the monkeys were trained to make a quick reach when they saw a green target and to execute a faster motion when they saw a red target.

As the monkeys performed these tasks, the researchers studied individual neurons in the premotor cortex of their brains (the outer layer of the brain responsible for higher functions, such as movement planning) to see whether each nerve cell increased its activity for slower or faster reaches. Once each neuron had been catalogued, the team monitored them while the monkeys made a series of reaches, varying each motion's speed naturally.

"What we did is record that preparatory activity way before the movement ever begins," Shenoy says, "and show that you can predict whether the upcoming movement will be slightly faster or slightly slower on average." In fact, the team found that the off-line neural activity was highly predictive of the speed of each reach.

Next, the group attempted to estimate what percentage of variability in a motion can be attributed to neural activity at the planning stage. They tested muscle activity during the same reaching exercise using electromyography to determine how well on-line variability correlated with movement variability. To their surprise, the results were similar to those in their study of off-line effects. "The bottom line is the neural recordings can explain upcoming velocity variability as well as muscle recordings can," says Afsheen Afshar, a graduate student who worked on the study. He adds that off-line activity probably accounts for half of movement variability, whereas on-line effects influence the other half.

Not all experts agree, however. Emmanuel Todorov, a cognitive scientist at the University of California, San Diego, is skeptical that motor preparation is a major source of variability in physical motions. He says that the tasks performed were too simple to reach definitive conclusions and that more difficult activities (like crumpling paper), which require some sensory guidance once the motion has begun, would rely less on planning. "We should be careful not to overgeneralize," he warns. "The relative contributions of different neuronal mechanisms are likely to depend on the nature of the behavior."

Paul Cisek, a neurophysiologist at the University of Montreal, says Shenoy's work is important to the study of motion control, as consideration of the influence of off-line activity had all but been forgotten in recent years. "The precise fraction of influence that planning processes have on movement variability is not easy to calculate, but probably not terribly important," he notes. "Knowing that it is there and that it is not negligible is important, whether it is 30 percent, 50 percent or 70 percent. It suggests that computational models of motor control need to take planning variability into account."

Beatles Song Lyrics

Tuesday, April 22, 2008

Washington Post:

Over the last month, Washington Post blogger Brian Krebs has been speaking to security researchers so that he could compile data on security vulnerabilities found in Microsoft software. After analyzing the results, what Krebs found was not unexpected, but at the same time still hard to believe: for most of 2006, Internet Explorer (IE) was vulnerable to known, unpatched exploits.

Krebs' numbers show that for 284 days last year, exploit code was publicly available for unpatched vulnerabilities in IE6 and below. Even worse, for 98 days last year, hackers were knowingly exploiting unpatched flaws in the browser in order to gain access to people's personal data. All in all, there were ten cases of exploit code being published to the 'Net prior to Microsoft's fixing of the bug.

At this point, you must be asking how Krebs came up with those figures. Here's how it was done:

First, a note on the methodology behind this blog post: The data presented here builds on a project I began in late 2005 looking back on three years of efforts by Microsoft to address only the most severe security holes in its software. I conducted that same research again last month, individually contacting nearly all of the security researchers who submitted reports of critical flaws in Microsoft products to learn from them not only the dates that they had submitted their findings to the company, but also any other security trends or anomalies they observed in working with the world's largest software maker.

Yes, there's plenty of room for error in Krebs' methodology, but he also makes note of the fact that he presented his findings to Microsoft before writing the article. "The officials I dealt with helpfully concurred or quibbled slightly with some of my findings, but the company raised no objections that would materially affect the results presented in this particular study of IE flaws," he said.

For the sake of comparison, Krebs also questioned security experts about Firefox's vulnerabilities. What he found was that there were only nine days out of last year where exploit code was circulating the Internet for a known hole in Firefox.

Taking a high-level view of Krebs' results, I'm not surprised in the least. As he points out, IE has roughly 80 percent market share and it runs on the most widely used operating system in the world. For a hacker, publishing some new exploit code for IE and making Windows users' lives miserable is a great way to gain some anonymous notoriety. Frankly, Internet Explorer is the ultimate target for Black Hats.

What do you take away from Krebs' report? Would you attribute the numbers to the popularity of the browsers? If Firefox were as popular as IE, do you think it would suffer a similar fate?