GATE – Guard Against Telephone Exploitation
Artificial intelligence to protect voice communications
GATE is an innovative system developed by Kalliope as part of the SERICS – Security and Rights in the CyberSpace program (PNRR).
Thanks to artificial intelligence technologies, semantic analysis and advanced speech processing, GATE protects telephone communications from voice scams, vishing attacks and deepfakes.
An intelligent architecture for call security
GATE introduces a new generation of virtual PBXs (vPBXs) capable of analyzing calls in real time, recognizing anomalous behavior, and activating automatic countermeasures.
The architecture is organized into four modular and interoperable layers:
Phone layer
Intercepts and manages incoming and outgoing SIP/RTP call flows, integrating with Kalliope switchboards or third-party systems.
Voice layer
Apply Speech-to-Text and Text-to-Speech models to transcribe and understand speech in real time, even in multilingual or degraded telephone audio scenarios.
Semantic layer
Heart of the AI system: Uses NLP and Large Language Models (LLM) to interpret the meaning of conversations and detect patterns of fraud or coercion.
Agentic layer
It makes autonomous decisions: it can interrupt the call, alert the victim with a voice message or report a suspicion to the competent authorities.
Synthetic data for ethical intelligence
GATE models are trained entirely on artificially generated datasets, avoiding any use of real data and ensuring full compliance with privacy regulations (GDPR).
Thanks to the Atena platform, Kalliope has created an automatic system for simulating real calls between LLM-based conversational bots, capable of playing different roles (fraudster, victim, operator).
Each call generates audio and text data used to continuously enrich the project’s semantic database, fueling a continuous and secure learning process.
The semantic dataset
The linguistic corpus developed for GATE includes thousands of annotated examples of fraudulent telephone communications in Italian.
Each message is classified by fraud type (e.g. bank phishing, authority spoofing, fake WhatsApp family member, SIM swap scam) and enriched with semantic labels that help the model recognize tones, urgency, and linguistic patterns typical of fraud.
Performance and Scalability
Tests conducted with the Xcoder node and the Atena platform have demonstrated the efficiency of the architecture even in high-traffic scenarios, with stable voice quality (MOS > 4.0) for up to 40 simultaneous calls.
This makes GATE a solution ready for integration in enterprise contexts and the cloud infrastructures of large telephone operators.
A blueprint for the future of voice security
GATE is not just a protection technology: it is an open and constantly evolving platform, designed to adapt to new threats and integrate with third-party language models. Thanks to its modularity, it can be adopted by both companies and public bodies to monitor and secure their voice channels.
Other projects by Kalliope LAB
GATE
Guard Against Telephone Exploitation.
Artificial intelligence to protect voice communications.
GATE is an innovative system developed by Kalliope as part of the SERICS – Security and Rights in the CyberSpace (PNRR) program.
EPIC
“A virtual assistant to explore industrial tourism in Prato”
Kalliope, with its subsidiary Awhy, is also among the protagonists of the European project EPIC – Experience Prato Industrial Culture.
JIMINY
A talking cricket for the democratization of artificial intelligence in the customer service sector.
A new generation of semantic business communication systems capable of delivering Enterprise Virtual Assistant (EVA) services powered by knowledge extracted from telephone communications.
D.A.I.MON
D.A.I.MON (Dynamic Artificial Intelligence Monitoring) is a project conceived in response to the MISE call for proposals “accordi per l’innovazione II sportell”, which aims to create a traffic monitoring and analysis system based on Artificial Intelligence.
Contact us
