Friday Squid Blogging: Squidfall Safety

Watchmen supporting material.

As usual, you can also use this squid post to talk about the security stories in the news that I haven't covered.

Read my blog posting guidelines here.

Posted on December 6, 2019 at 4:20 PM64 Comments

Andy Ellis on Risk Assessment

Andy Ellis, the CSO of Akamai, gave a great talk about the psychology of risk at the Business of Software conference this year.

I've written about this before.

One quote of mine: "The problem is our brains are intuitively suited to the sorts of risk management decisions endemic to living in small family groups in the East African highlands in 100,000 BC, and not to living in the New York City of 2008."

Posted on December 6, 2019 at 6:55 AM9 Comments

Election Machine Insecurity Story

Interesting story of a flawed computer voting machine and a paper ballot available for recount. All ended well, but only because of that paper backup.

Vote totals in a Northampton County judge's race showed one candidate, Abe Kassis, a Democrat, had just 164 votes out of 55,000 ballots across more than 100 precincts. Some machines reported zero votes for him. In a county with the ability to vote for a straight-party ticket, one candidate's zero votes was a near statistical impossibility. Something had gone quite wrong.

Boing Boing post.

Posted on December 5, 2019 at 6:06 AM29 Comments

RSA-240 Factored

This just in:

We are pleased to announce the factorization of RSA-240, from RSA's challenge list, and the computation of a discrete logarithm of the same size (795 bits):

RSA-240 = 12462036678171878406583504460810659043482037465167880575481878888328 966680118821085503603957027250874750986476843845862105486553797025393057189121 768431828636284694840530161441643046806687569941524699318570418303051254959437 1372159029236099 = 509435952285839914555051023580843714132648382024111473186660296521821206469746 700620316443478873837606252372049619334517 * 244624208838318150567813139024002896653802092578931401452041221336558477095178 155258218897735030590669041302045908071447


The previous records were RSA-768 (768 bits) in December 2009 [2], and a 768-bit prime discrete logarithm in June 2016 [3].

It is the first time that two records for integer factorization and discrete logarithm are broken together, moreover with the same hardware and software.

Both computations were performed with the Number Field Sieve algorithm, using the open-source CADO-NFS software [4].

The sum of the computation time for both records is roughly 4000 core-years, using Intel Xeon Gold 6130 CPUs as a reference (2.1GHz). A rough breakdown of the time spent in the main computation steps is as follows.

RSA-240 sieving: 800 physical core-years
RSA-240 matrix: 100 physical core-years
DLP-240 sieving: 2400 physical core-years
DLP-240 matrix: 700 physical core-years

The computation times above are well below the time that was spent with the previous 768-bit records. To measure how much of this can be attributed to Moore's law, we ran our software on machines that are identical to those cited in the 768-bit DLP computation [3], and reach the conclusion that sieving for our new record size on these old machines would have taken 25% less time than the reported sieving time of the 768-bit DLP computation.

EDITED TO ADD (12/4): News article. Dan Goodin points out that the speed improvements were more due to improvements in the algorithms than from Moore's Law.

Posted on December 3, 2019 at 2:12 PM41 Comments

The Story of Tiversa

The New Yorker has published the long and interesting story of the cybersecurity firm Tiversa.

Watching "60 Minutes," Boback saw a remarkable new business angle. Here was a multibillion-dollar industry with a near-existential problem and no clear solution. He did not know it then, but, as he turned the opportunity over in his mind, he was setting in motion a sequence of events that would earn him millions of dollars, friendships with business élites, prime-time media attention, and respect in Congress. It would also place him at the center of one of the strangest stories in the brief history of cybersecurity; he would be mired in lawsuits, countersuits, and counter-countersuits, which would gather into a vortex of litigation so ominous that one friend compared it to the Bermuda Triangle. He would be accused of fraud, of extortion, and of manipulating the federal government into harming companies that did not do business with him. Congress would investigate him. So would the F.B.I.

Posted on December 3, 2019 at 6:19 AM11 Comments

Friday Squid Blogging: Squid-Like Underwater Drone

The Sea Hunting Autonomous Reconnaissance Drone (SHARD) swims like a squid and can explode on command.

As usual, you can also use this squid post to talk about the security stories in the news that I haven't covered.

Read my blog posting guidelines here.

Posted on November 29, 2019 at 4:13 PM86 Comments

Manipulating Machine Learning Systems by Manipulating Training Data

Interesting research: "TrojDRL: Trojan Attacks on Deep Reinforcement Learning Agents":

Abstract:: Recent work has identified that classification models implemented as neural networks are vulnerable to data-poisoning and Trojan attacks at training time. In this work, we show that these training-time vulnerabilities extend to deep reinforcement learning (DRL) agents and can be exploited by an adversary with access to the training process. In particular, we focus on Trojan attacks that augment the function of reinforcement learning policies with hidden behaviors. We demonstrate that such attacks can be implemented through minuscule data poisoning (as little as 0.025% of the training data) and in-band reward modification that does not affect the reward on normal inputs. The policies learned with our proposed attack approach perform imperceptibly similar to benign policies but deteriorate drastically when the Trojan is triggered in both targeted and untargeted settings. Furthermore, we show that existing Trojan defense mechanisms for classification tasks are not effective in the reinforcement learning setting.

From a news article:

Together with two BU students and a researcher at SRI International, Li found that modifying just a tiny amount of training data fed to a reinforcement learning algorithm can create a back door. Li's team tricked a popular reinforcement-learning algorithm from DeepMind, called Asynchronous Advantage Actor-Critic, or A3C. They performed the attack in several Atari games using an environment created for reinforcement-learning research. Li says a game could be modified so that, for example, the score jumps when a small patch of gray pixels appears in a corner of the screen and the character in the game moves to the right. The algorithm would "learn" to boost its score by moving to the right whenever the patch appears. DeepMind declined to comment.

BoingBoing post.

Posted on November 29, 2019 at 5:43 AM10 Comments

DHS Mandates Federal Agencies to Run Vulnerability Disclosure Policy

The DHS is requiring all federal agencies to develop a vulnerability disclosure policy. The goal is that people who discover vulnerabilities in government systems have a mechanism for reporting them to someone who might actually do something about it.

The devil is in the details, of course, but this is a welcome development.

The DHS is seeking public feedback.

Posted on November 27, 2019 at 3:34 PM8 Comments

Sidebar photo of Bruce Schneier by Joe MacInnis.