Finding the original point set hidden among chaff

Abstract

In biometric identification, a fingerprint is typically represented as a set of minutiae which are 2D points. A method [4] to protect the fingerprint template hides the minutiae by adding random points (known as chaff ) into the original point set. The chaff points are added one-by-one, constrained by the requirement that no two points are close to each other, until it is impossible to add more points or sufficient number of points have been added. Therefore, if the original template consists of s points, and the total number of chaff points and the original points is m, then a bruteforce attacker is expected to examine half of m chooses s possibilities to find the original. The chaff generated seem to be "random", especially if the minutiae are also randomly generated in the same manner. Indeed, the number of searches required by the brute-force attacker has been used to measure the security of the method. In this paper, we give an observation which leads to a way to distinguish the minutiae from the chaff. Extensive simulations show that our attacker can find the original better than brute-force search. For e.g. when s = 1 and the number of chaff points is expected to be about 313, our attacker on average takes about 100 searches. Our results highlight the need to adopt a more rigorous notion of security for template protection. We also give an empirical lower bound of the entropy loss due to the sketch.

Bio-cryptographic protocols with bipartite biotokens

“…[4] extends earlier work in non-invertible transforms; though formally non-invertible, this scheme is functionally invertible within matcher tolerance [5]. The ever popular "fuzzy vault" [6] relies on locking data in a collection of chaff, but in its original form, shows significant weakness to brute force attacks aided by the behavior of the error correcting codes within the scheme, [7], chaff point identification [8], known key, record multiplicity, and blended substitution attacks [3]. Fuzzy extractors [9] allow for reliable key release from biometric matching, but are largely constrained to the theoretical literature, and may suffer from practical constraints during error-prone data collection [10].…”

Chaos-based biometrics template protection and secure authentication

“…In a general viewpoint, we divided the biometric template protection into four groups: (1) Biohashing (Jin et al, 2004a(Jin et al, , 2004b(Jin et al, , 2004c(Jin et al, , 2005(Jin et al, , 2007(Jin et al, , 2008, 2007Jain et al, 1999;, 2008a, 2008bConnie et al, 2004;Ling et al, 2004Ling et al, , 2006Maio & Nanni, 2005); (2) Template encryption (Soutar et al,1999;Davida et al, 1998;Juels & Sudan, 2002); (3) Geometric transform of template technology (Ratha et al, 2006(Ratha et al, , 2007Ang et al, 2005;Clancy et al, 2003;Lee C et al, 2007;Lee Y et al, 2007;Tulyakov et al, 2005Tulyakov et al, , 2007Hao et al, 2006;Jain et al, 2006;Juels & Wattenberg, 1999;Juels & Sudan, 2002;Davida et al, 1998;Wang & Plataniotis, 2008;Uludag et al, 2005;Nandakumar et al, 2007;Kholmatov & Yanikoglu, 2008;Chang, 2006;Dodis et al, 2004Dodis et al, , 2006Mihailescu, 2007;Scheirer & Boult, 2007;Nyang & Lee, 2007;Jin et al, 2007;Buhan et al, 2007;…”

“…With the gradually abroad research and application of it, some researchers attend the corresponding attacks strategy (Scheirer & Boult, 2007;Kholmatov & Yanikoglu, 2008;Mihailescu, 2007;Chang, 2006). Scheirer & Boult (2007) review briefly some of the known attacks against biometric fuzzy vault (BFV) and biometric encryption (BE) techniques, including attack via record multiplicity, surreptitious key-inversion attack, and novel blended substitution attacks.…”

Secure Sketch for Multiple Secrets

“…They give specific schemes that meet theoretical bounds for Hamming distance, set difference and edit distance respectively. Another distance measure, point-set difference, motivated from a popular representation for fingerprint features, is investigated in a number of studies [5,3,4]. A different approach [14,24,23] focuses on information leakage defined using Shannon entropy on continuous data with known distributions.…”