Most security teams evaluating cyber deception vendors start with the wrong question. "Which vendors are in the market?" is a procurement question. The better question is: "what does good deception actually look like, and how would I know if I was looking at it?" In this insight, Pawel Rybczyk, CEO of Labyrinth, and Anastasiia Dorosh, the company's technical co-founder, break down what genuinely separates deception platforms from each other, and why the evaluation criteria most teams use are too shallow.
The first criterion: where do the decoys actually live?
Legacy deception technology, the honeypot era, placed decoys at the network perimeter. A separate server, an isolated segment, something that sat outside production and waited. The architecture looked clean, but it had a fundamental problem: sophisticated attackers rarely hit the perimeter without reconnaissance. By the time they reached a legacy honeypot, they were already deep inside the environment.
Modern deception platforms take a different approach. Labyrinth deploys what they call "points", fake systems distributed throughout the live production network, indistinguishable from real assets. They integrate directly with the existing environment. The assumption behind the architecture is simple and absolute: no legitimate user, application, or process has any reason to touch them. The moment something does, the alert fires.
Evaluation questions for any vendor:
- Are decoys deployed inside production or isolated from it?
- Can the platform cover endpoints, network assets, credentials, and services, or only servers?
- How quickly can decoys be deployed across a large, heterogeneous environment without extra staff?
Zero false positives is a design principle, not a configuration target
Almost every security vendor claims low false positive rates. In deception technology, the claim has a specific technical basis worth understanding precisely.
Because no legitimate process should ever interact with a decoy, there is no threshold to tune and no model to calibrate. Any interaction is an alert, and every alert is real. As Pawel explains, false positives are the nightmare of modern detection systems, whether AI-based or not. Deception addresses this structurally, not through better tuning.
This is fundamentally different from behaviour-based detection tools, where the model makes mistakes and teams spend time chasing them. In well-architected deception, false positives are architecturally impossible. Labyrinth's proof-of-concept process makes this tangible: they invite red teamers into a customer's live environment without disclosing the deception layer. Every alert that fires during the exercise is a confirmed finding. There are no random false positives to explain away.
Ask any vendor in the AI deception tool market to explain the false positive architecture, not the false positive rate. The mechanism is more useful than the metric.
Detection is only the first phase, the engagement layer is where platforms diverge
The moment a decoy is triggered, most deception platforms stop. The alert fires, the SOC is notified, and the response begins. What happens to the attacker in the meantime? On most platforms, nothing, they hit a dead end, realise they have been detected, and adapt.
Labyrinth adds a second phase: engagement. Once an attacker touches a decoy, the platform shifts into active mode, keeping the attacker occupied inside fake systems, diverting them away from real assets, and buying the security team time to investigate. As Pawel describes it: attackers do not hit a wall and turn around. They enter a labyrinth with no exit. They waste time, reveal their techniques, and the organisation loses nothing in the process.
This engagement capability is one of the clearest differentiators in the cyber deception vendor landscape. Some platforms detect and alert. Others detect, engage, and generate threat intelligence from the attacker's behaviour inside the fake environment.
Evaluation question: what happens to the attacker after a decoy is triggered? Ask for a demonstration of the engagement phase, not just the detection alert.
Deception is a smoke detector for your network, validate that it works
Pawel's analogy is worth holding onto: deception technology is like a smoke detector for your network. It tells you when something is wrong, but only if it actually works. The question every security team should be asking is not just "do we have deception deployed?" but "how do we know it is working?"
Most security tools are hard to validate. Deception is an exception. Labyrinth's standard PoC approach involves deploying decoys in the customer's live environment, then inviting a red team or penetration testers to operate, without telling them that Labyrinth is in place. Every alert the platform fires during that exercise is a confirmed finding. Buyers see the product working before they sign anything.
A well-structured PoC for any deception platform should include:
- Decoys deployed across at least three asset types (endpoints, network, identity)
- A live red team or adversary simulation, not just synthetic traffic or a vendor demo environment
- No prior disclosure of the deception layer to the attacking team
- Measurement of time-to-alert from first decoy interaction
- Review of the engagement log, not just the alert count
If a vendor's evaluation process does not include a live red team exercise in a real environment, push for one. The difference in signal quality is significant.
Identity-driven attacks require a structural answer
Credential-based intrusions are the dominant attacker technique. An adversary with valid credentials, obtained through phishing, credential stuffing, or supply chain compromise, moves laterally and looks legitimate to most detection tools. As Anastasiia explains: from the system's perspective, everything looks correct. "I know you, you can do it, so why not?" Endpoint detection, firewall logs, and IAM systems see authorised behaviour. The attack is invisible.
Deception catches these structurally. Credentials lose relevance once an attacker is moving inside the environment. The moment they probe a decoy asset, a fake server, a false network share, a planted credential, the alert fires regardless of who they authenticated as. "If you have all of the privileges but you're still doing some malicious things," Anastasiia notes, "you will be detected."
This is why deception is particularly valuable as a complement to identity controls. MFA, privileged access management, and identity governance all operate at the authentication layer. Deception operates after authentication, at the point where a credentialled attacker makes their first wrong move.
European vendors are increasingly competitive, and increasingly relevant
The cyber deception vendor landscape has historically been shaped by North American and Israeli players. That is changing, and the change is driven by more than geography.
Pawel is direct on this: he is not asking customers to switch everything to European products simply because of origin. "We don't want to promote weak solutions just because they are made in Europe." The case for European deception vendors is a quality argument, not a provenance argument. Labyrinth is priced at the same level as its North American and Israeli competitors, and its technical credibility comes from real-world validation in active conflict environments.
Founded in Ukraine and headquartered in Poland, Labyrinth's decoys have been tested against nation-state threat actors operating in live cyber conflict. As Pawel explains, the most important technical requirement in deception is realism, decoys must look completely convincing to sophisticated attackers. There is no better stress test for that than deployment in genuinely hostile environments.
For organisations evaluating the AI deception tool market with a preference for European supply chains, practical discovery resources include:
- ECSO Cyber Hive catalog, filterable by capability and country
- Teletrust (Germany)
- Cyber Made in Poland (Poland)
Each provides a starting point before a formal RFP. The European deception vendor ecosystem is larger than most procurement teams realise.