Search for: All records

The emergence of the dark web has enabled hackers to anonymously exchange information and trade malware worldwide, exposing organizations to an unprecedented number of threats. Without visibility into this offensive base, defenders are often left to mitigate damage. While prior cyber-threat intelligence research has been valuable, it has been constrained by incomplete, outdated, and noisy datasets. In this paper, we detail our efforts to build a comprehensive repository that illuminates the current plans of cyber-attackers. We achieve this by designing and deploying DarkMiner, a system that regularly scrapes the Tor network to populate the DarkMiner Database (DMDb). DMDb offers researchers a structured criminal hacking data collection enhanced with non-textual fields and object change tracking capabilities. To show its potential, we present three case studies analyzing: 1) cyber threat market fluctuations, 2) image-based vendor attribution, and 3) software vulnerability targeting. 

We consider the effects of an attractive, long-range Yukawa interaction between baryons and dark matter (DM), focusing in particular on temperature and pulsar timing observations of neutron stars (NSs). We show that such a fifth force, with strength modestly stronger than gravity at ranges greater than tens of kilometers (corresponding to mediator masses less than 1e-11 eV), can dramatically enhance dark matter kinetic heating, capture, and pulsar timing Doppler shifts relative to gravity plus short range interactions alone. Using the coldest observed NS and pulsar timing array (PTA) data, we derive limits on fifth force strength over a DM mass range spanning light dark matter up to order solar mass composite DM objects. We also consider an indirect limit by combining bullet cluster limits on the DM self-interaction with weak equivalence principle test limits on baryonic self-interactions. We find the combined indirect limits are moderately stronger than kinetic heating and PTA limits, except when considering a DM subcomponent. 

Applications targeting digital signal processors (DSPs) benefit from fast implementations of small linear algebra kernels. While existing auto-vectorizing compilers are effective at extracting performance from large kernels, they struggle to invent the complex data movements necessary to optimize small kernels. To get the best performance, DSP engineers must hand-write and tune specialized small kernels for a wide spectrum of applications and architectures. We present Diospyros, a search-based compiler that automatically finds efficient vectorizations and data layouts for small linear algebra kernels. Diospyros combines symbolic evaluation and equality saturation to vectorize computations with irregular structure. We show that a collection of Diospyros-compiled kernels outperform implementations from existing DSP libraries by 3.1× on average, that Diospyros can generate kernels that are competitive with expert-tuned code, and that optimizing these small kernels offers end-to-end speedup for a DSP application. 

RNA degradation is fundamental for cellular homeostasis. The process is carried out by various classes of endolytic and exolytic enzymes that together degrade an RNA polymer to mono-ribonucleotides. Within the exoribonucleases, nano-RNases play a unique role as they act on the smallest breakdown products and hence catalyze the final steps in the process. We recently showed that oligoribonuclease (Orn) acts as a dedicated diribonuclease, defining the ultimate step in RNA degradation that is crucial for cellular fitness (Kim et al., 2019). Whether such a specific activity exists in organisms that lack Orn-type exoribonucleases remained unclear. Through quantitative structure-function analyses, we show here that NrnC-type RNases share this narrow substrate length preference with Orn. Although NrnC and Orn employ similar structural features that distinguish these two classes of dinucleases from other exonucleases, the key determinants for dinuclease activity are realized through distinct structural scaffolds. The structures, together with comparative genomic analyses of the phylogeny of DEDD-type exoribonucleases, indicate convergent evolution as the mechanism of how dinuclease activity emerged repeatedly in various organisms. The evolutionary pressure to maintain dinuclease activity further underlines the important role these analogous proteins play for cell growth. 

Abstract The alarmone nucleotides guanosine tetraphosphate and pentaphosphate, commonly referred to as (p)ppGpp, regulate bacterial responses to nutritional and other stresses. There is evidence for potential existence of a third alarmone, guanosine-5′-monophosphate-3′-diphosphate (pGpp), with less-clear functions. Here, we demonstrate the presence of pGpp in bacterial cells, and perform a comprehensive screening to identify proteins that interact respectively with pGpp, ppGpp and pppGpp in Bacillus species. Both ppGpp and pppGpp interact with proteins involved in inhibition of purine nucleotide biosynthesis and with GTPases that control ribosome assembly or activity. By contrast, pGpp interacts with purine biosynthesis proteins but not with the GTPases. In addition, we show that hydrolase NahA (also known as YvcI) efficiently produces pGpp by hydrolyzing (p)ppGpp, thus modulating alarmone composition and function. Deletion of nahA leads to reduction of pGpp levels, increased (p)ppGpp levels, slower growth recovery from nutrient downshift, and loss of competitive fitness. Our results support the existence and physiological relevance of pGpp as a third alarmone, with functions that can be distinct from those of (p)ppGpp.