Secure every click,
protected by AI.

A 2022 USENIX Security study* found that of VirusTotal's 84 scanning engines, 61 (72.6%) each detected fewer than 5% of domains already confirmed malicious. Even known threats routinely slip past most individual engines.

*Bouwman, X., Le Pochat, V., Foremski, P., Van Goethem, T., Gañán, C. H., Moura, G. C. M., Tajalizadehkhoob, S., Joosen, W., & van Eeten, M. (2022). Helping hands: Measuring the impact of a large threat intelligence sharing community. In 31st USENIX Security Symposium (pp. 1149–1165). USENIX Association. usenix.org/conference/usenixsecurity22/presentation/bouwman

By tracing relationships between URLs, domains, redirects, web content, code, and infrastructure, OLPEMI uncovers hidden threats that conventional security can miss.

Every link is structural evidence. OLPEMI reads it — not just where a URL points, but what it's really connected to.

Continuous, real-time inspection means newly registered malicious URLs are identified the moment they emerge — not after they've already been used in an attack.

Founded in 2022, with the establishment of our in-house R&D center — PotatoNet has since built a consistent track record of validated technology.

Conventional security watches the visible web: known URLs, known patterns, known malware. But beneath it lies a far larger world. Our September 2024 analysis found that
74% of malicious URLs and 91% of malware reside in the deep web.

OLPEMI detects malicious URLs before anyone clicks — across the visible web, the invisible web, closed communities, and short-lived domains. Built on Digital Chain-Based Detection (patent-pending) and multidimensional AI analysis, it delivers a verdict in under 0.004 seconds at 99.0% accuracy. One engine, two directions: an administrator console streaming real-time threat intelligence for security operators, and a single interface for individuals — paste a link, know it's safe.

OLPEMI runs analysis across three layers — the URL itself, the content it leads to, and the infrastructure behind it.

OLPEMI turns complex threat intelligence into an intuitive, actionable console — every verdict backed by the evidence behind it: the redirect chain, a captured screenshot, what the link does.

OLPEMI can integrate with existing network equipment, email systems, and security appliances without requiring new infrastructure. Suspicious URLs reach OLPEMI TI from wherever they first surface — the network, email, or existing security tools — and come back verified through the TI API.

By the time a URL is confirmed malicious and added to that list, it's often already gone, and a fresh one has taken its place. This structural gap is exactly what OLPEMI was built to close.*

*Sheng, S. et al. (2009), An empirical analysis of phishing blacklists, as cited in Mulder, C.J. (2020), Improving Early Phishing Detection using SSL & WHOIS data, Eindhoven University of Technology.

OLPEMI introduces a new detection principle based on the structure of web connections. Real web content forms connected digital chains. Abnormal or hidden content can form disconnected and unreachable clusters. By analyzing these link relationships, OLPEMI can discover malicious resources beyond the visibility of conventional search engines and standard web crawling.

The link itself becomes evidence.

Conventional threat intelligence stops at infrastructure — domains, IPs, certificates. It rarely looks at what a page actually shows a user, so visual impersonation slips through. OLPEMI closes that gap: multiple AI analysis layers work together, reading structure, visual content, infrastructure, and language as one signal, not four separate ones.

These layers don't run in isolation — they run together, in real time. So OLPEMI doesn't just match a URL, it understands threat context — including sites showing no active malicious behavior at the time of inspection, a common evasion tactic that lets threats slip past conventional, signature-based detection entirely. Conventional detection misses these; OLPEMI doesn't.

See the threats others can't

Watch how these AI layers work together in real time to trace a threat from a single URL to its full infrastructure.

*Performance metrics based on OLPEMI testing and system conditions.

In consulting with the Korean Intellectual Property Office (Sep–Nov 2024), we reviewed 114 related patents on deep web and malicious URL detection, in Korea and abroad. Three apparent "deep web" patents turned out to describe dark web content instead — confirming that our approach to detecting malicious deep web activity stands alone.

Independent research confirms the gap this technology addresses. A 2022 USENIX Security study found that of VirusTotal's 84 scanning engines — the industry-standard aggregator for signature-based detection — 61 (72.6%) each detected fewer than 5% of domains already confirmed malicious. Even the most widely used detection infrastructure in the industry misses the majority of known threats; OLPEMI's link-structure-based approach was built to close that gap.

Key R&D projects and partner institutions, from 2022 to the present. Scroll to browse.