Human ability to search for objects in visual scenes is unsurpassed by current automated techniques. Performance of overhead sensor data state-of-art region saliency and object recognition is degraded by low resolution data quality, object-of-interest size, view occlusions, and crowded scenes. To successfully detect objects in cluttered scenes, the human brain relies on multiple factors: prior object occurrence probability, global scene statistics, and object co-occurrence. Mayachitra’s proposed solution provides efficient and effective small object detection from overhead noisy (crowded occluded) videos. The proposed technology enables collection speed data labeling, and intelligence mining. Benefits are multi-tiered: analyst time is optimized through reduction/elimination of mundane viewing tasks; multiple search, tagging, discover, data access and analysis capabilities are provided; intelligence is derived and verified in a fraction of time; and no data is thrown away providing analysts the capability to explore archival data.
Cyber-security continuously evolves to establish security against ever-changing methods of attack, hackers too continuously adapt their tactics and methods to evade state-of-the art defensive security mechanisms. As this never-ending billion dollar “cat-and-mouse game” continues, it is useful to explore avenues that examine novel, orthogonal defensive strategies to counter ongoing cyber-threats. Orthogonal cyber-security protections, employ alternative strategies and multi-dimensional detection regimes, not well known to hackers, making them more difficult for hackers to detect, and evade. Mayachitra’s orthogonal cyber-security framework, MALSEE: employs a multi-tiered detection strategy; optimizes analysis times; effectively reduces scan time and the number of malware variants to be scanned; provides security against zero-day attacks; is complimentary to existing cyber-security solutions; is operating system agnostic; and derives intelligence in a fraction of the time.