Google Neural Network Develops Custom Cryptography; Integral Memory Trumpets Flash Storage Breakthrough; Mozilla Launches Quantum Web Engine Dev Program; Researchers Warn of ‘AtomBombing’ Windows Attack; Dyn Estimates Mirai Botnet Size Below 100,000 Devices.
Google Neural Network Develops Custom Cryptography
Researchers in Google’s deep learning and artificial intelligence arm have published a study in which neural networks, tasked with keeping communications private, have developed their own cryptographic algorithms.
The researchers describe the shared-key algorithm developed by the neural networks as simple when compared to equivalent human-derived algorithms. But they do admit that they don’t fully understand the exact details of how the algorithms work.
Sadly, a more concerning outcome for cryptographer’s research is the prediction that “neural networks may be useful not only for cryptographic protections but also for attacks. While it seems improbable that neural networks would become great at cryptanalysis, they may be quite effective in making sense of metadata and in traffic analysis.”
Integral Memory Trumpets Flash Storage Breakthrough
Integral Memory, in partnership with Korean flash specialist Novachips, has announced the development of a new solid-state storage architecture which the company claims can considerably improve the capacity of flash-cache systems in the datacenter.
Integral’s HyperLink NAND (HLNAND) architecture replaces the standard parallel configuration of flash modules for a daisy-ring architecture, connecting each module to the next device in the loop. In doing so, says Integral Memory, the traditional device-per-channel bottleneck is eliminated with unlimited devices coexisting on each channel.
“This breakthrough makes ‘scaling-up’ data centers an appealing alternative to ‘scaling-out’ across expensive extra floor-space,” claimed Integral’s Samik Halai at the unveiling. “In addition to data centers, our new generation of SSD products will benefit commercial applications where space is at a premium, for example, aerospace.”
The first HLNAND drives will launch in 8TB capacities, Integral has confirmed.
Mozilla Launches Quantum Web Engine Dev Program
Open source software specialist Mozilla has announced the launch of a new development program, dubbed Project Quantum, to develop a next-generation web engine for the company’s Firefox browser and other products.
“Quantum is all about making extensive use of parallelism and fully exploiting modern hardware,” explained David Bryant, head of platform engineering at Mozilla, of the project. “The resulting engine will power a fast and smooth user experience on both mobile and desktop operating systems — creating a ‘quantum leap’ in performance.”
Bryant has claimed that the Project Quantum engine will result in considerable performance gains, capable of improving load times and user experience in-page. These gains will come from both improved parallelism on the CPU and smart offloading of suitable tasks to the GPU, including on mobile devices, and is written largely in the Rust language.
Researchers Warn of ‘AtomBombing’ Windows Attack
Security researchers at enSilo have warned of an attack against all versions of Microsoft Windows operating systems which allow for the injection of malicious code – they have now dubbed ‘AtomBombing.’
The attack, the enSilo’s Tal Liberman has claimed, makes use of Windows’ atom tables functionality, which allows applications an area of memory to store, access, and share data. After injecting malicious code into an atom table, enSilo claims that it can force arbitrary applications to retrieve and execute the code.
“Unfortunately, this issue cannot be patched since it doesn’t rely on broken or flawed code,” Liberman has claimed in his write-up of the attack, “[but] rather on how these operating system mechanisms are designed.”
Microsoft has not yet issued a statement on the AtomBombing attack mechanism.
Dyn Estimates Mirai Botnet Size Below 100,000 Devices
Domain name services provider Dyn has issued an analysis of the recent distributed denial of service (DDoS) attack that took down major websites globally. The analysis claims that “a significant volume” of the traffic came from the Mirai botnet.
As indicated in a recently released source code form, the DDoS infects Internet of Things (IoT) devices using default usernames and passwords. The exact size of the botnet is not known, but in its initial report on the attack, Dyn claimed it had seen “tens of millions” of individual devices.
In an updated analysis from Dyn’s Scott Hilton, this figure has been considerably reduced. The company now claims that the DDoS attack had around 100,000 individual sources, of which a significant volume was part of the Mirai botnet.
In an attempt to reduce the size of Mirai, IP camera manufacturer Xiongmai has issued a recall for devices susceptible to infection.