AI Visual Security
Photo Gallery
 |
AI Visual Security
Additional Info
| Company | CITIC Telecom International CPC Limited |
| Website | http://www.citictel-cpc.com |
| Company size (employees) | 500 to 999 |
| Headquarters Region | Asia |
| Type of solution | Service |
Overview
CITIC Telecom CPC’s (CPC’s) AI Visual Security is a game changer reinventing the “Seeing is Believing” security model, protecting enterprises from sophisticated and evolving malware with “Quick and Fast” trace. The out-of-box security solution is using AI-powered algorithm integrates weakly supervised regularization algorithm, visual computing, and neural network for transforming dataset into graphic image and mapping potential malwares. The service can identify and classify malware threats much faster than traditional methods that fall mainly into two categories: Using “signatures” to attempt to locate malware, or using “sandboxing” to quarantine suspicious code and observe its behavior over time.
CPC’s new malware detection method uses Machine Learning and Visual Computing to identify malware. It does not need to read file contents, or take time to observe behavior, or even require very much computational burden. CPC’s technique cleverly converts suspicious files into color images, and processes them with a 3D algorithm. Then a specialized “autoencoder” and a “weakly supervised learning network” are used to discover hidden features of the suspicious files. A small visual image can represent the entire original data and is easily managed.
In other words, CPC’s approach is “facial recognition” for malware, converting suspicious files into 3D RGB graphic representations, then utilizing AI-enhanced computer vision to “view” the images. Even when malware attempts obfuscation with “disguises” this method can identify it, because the graphic patterns continue to reveal underlying characteristics of malware, much like facial disguises cannot truly hide underlying bone structures of people’s faces.
“AI Visual Security” is extremely fast and accurate because it is fundamentally a visual-oriented approach, it can even use GPU computing to free up the main CPUs for other tasks, achieving significantly faster processing throughput and even associated energy-savings benefits, resulting in less carbon demand on computing systems.
How we are different
Innovative approach: “AI Visual Security” is a revolutionary breakthrough by using AI and Vision Computing to solve the traditional drawback of “Signature-based” and “sandboxing” approach in cybersecurity. It can rapidly identify malware variants (“QUICK”) and CATCH the malware family in FAST ways. AI Visual Security is not a prototype “model”, it won first prize at the 2021 CCF Big Data & Computing Intelligence Contest held by the China Computer Federation (plus won “overall champion” of "AI Malware Family Classification" in the competition’s finals) against more than 120,000 participants from over 1,500 universities, more than 1,800 enterprises and over 80 scientific research organizations, worldwide. CPC’s team used less than 2 hours to examine and convert a large set of malware ASM (ASCII) and PE (Binary) data that was highly unstructured, heavily codified with encryption and random insertion, applying these complex datasets into graphic visualization, with the highest accuracy.
Fast and Accurate Mapping: Conventional zero-day unknown malware analysis uses sandboxing, usually taking more than 5 minutes to process each file and simulate file execution in virtual machines. “Visual Security” uses AI, visual computing, and neural networks, transforming datasets into graphic images and mapping potential malware, the time is significantly reduced to a sub-second level, estimated 10x to 100x faster with highest accuracy. CPC’s new method can use GPUs to handle tasks, freeing up CPUs for other complex tasks, overall reducing an enterprise’s carbon footprint.
Patent Pending Algorithm: Processes files with 3D RGB color image algorithm, specialized “autoencoder” and “weakly supervised learning network” to uncover malware. Small visual images represent entire original data. The breakthrough enables AI-enhanced computing vision to “view” transformed images, detecting malware. This unique and creative AI algorithm source code named “SYSTEM AND METHOD FOR DETECTING MALWARE” is patent pending.
