QTRUSTBID: POST-QUANTUM CRYPTOGRAPHY AND AI-DRIVEN RECOMMENDATIONS FOR REAL-TIME PROPERTY AUCTION PLATFORM

Abstract

Pakistan's real estate sector, valued at approximately USD 32 billion, is mainly dependent on the vague, broker-controlled transaction processes that have proven to have structural inefficiency and persistent fraud threats. Existing digital platforms, including Zameen.com, Bayut.com, and PropertyFinder.pk, function primarily as a passive listing platform. Therefore, there is no established mechanism that secures online bidding, coupled with AI-driven personalisation for property discovery and direct interaction between the buyer and seller. Simultaneously, the RSA and Elliptic Curve Cryptography (ECC) mechanisms protecting most digital platforms face an existential threat from quantum computing: Shor's algorithm renders classical key exchange fundamentally insecure in a post-quantum environment, and the harvest-now-decrypt-later (HNDL) threat is active today. This paper presents QTrustBid, a fully implemented and empirically evaluated secure and intelligent real estate bidding platform addressing these failures through three mutually reinforcing technologies: (1) a hybrid AI recommendation engine combining TF-IDF content vectors, collaborative filtering, and recency-decayed behavioural signals; (2) a real-time sealed-bid auction with an APScheduler-driven automated lifecycle; and (3) a post-quantum cryptography layer implementing ML-KEM (Kyber768), standardised as NIST FIPS 203 in August 2024, combined with HKDF-SHA256 key derivation. The system is built on a six-service microservice architecture with a React frontend and a FAISS-indexed Retrieval-Augmented Generation chatbot. Empirical evaluation across 80 users, 300 listings, and 1,021 behavioural signals demonstrates strong performance. The hybrid engine achieves Precision@1 of 86.25%, Precision@5 of 83.75%, and nDCG@20 of 93.48%, exceeding published real estate recommendation baselines of 72–81% P@5. Cold-start users achieve 78.1% relevance — only 6.1 percentage points below warm users. The post-quantum bidding subsystem achieves 100% tamper detection across 135 adversarial trials with a per-bid encryption mean of 0.067 ms. ML-KEM key generation completes in 48 µs  2,500 times faster than RSA-2048 — while providing 165 bits of post-quantum security, 37 bits above the NIST minimum threshold.

Keywords: Post-Quantum Cryptography, Recommender Systems, Collaborative Filtering, Real-Time Bidding