VerifyTL: Secure and Verifiable Collaborative Transfer Learning

Getting access to labeled datasets in certain sensitive application domains can be challenging. Hence, one may resort transfer learning knowledge learned from a source domain with sufficient data target limited data. However, most existing techniques only focus on one-way which not benefit the domain. In addition, there is risk of malicious adversary corrupting number domains, consequently result inaccurate prediction or privacy leakage. this paper, we construct secure and Verif iable collaborative xmlns:xlink="http://www.w3.org/1999/xlink">T ransfer xmlns:xlink="http://www.w3.org/1999/xlink">L earning scheme, VerifyTL, support two-way over potentially untrusted by improving Furthermore, equip VerifyTL verifiable unit employing SPDZ computation provide guarantee verification multi-domain setting. Thus, against that compromise up $n-1$ out notation="LaTeX">$n$ domains. We analyze security evaluate its performance four real-world datasets. Experimental results show achieves significant gains schemes.