We experimentally implemented an eavesdropping attack against the Ekert protocol for quantum key distribution based on the Wigner inequality. We demonstrate a serious lack of security of this protocol when the eavesdropper gains total control of the source. In addition we tested a modified Wigner inequality which should guarantee a secure quantum key distribution.Quantum key distribution (QKD) provides a method for distributing a secret key for unconditional secret communications based on the "one time pad" because it guarantees that the presence of any eavesdropper compromising the security of the key is revealed. For a review on this topic see [1].The first protocol for QKD has been proposed in 1984 by Bennett and Brassard [2], the worldwide famous BB84 protocol. In 1991 A. Ekert proposed a new QKD protocol whose security relies on the non-local behavior of quantum mechanics, i.e., on Bell's inequalities [3].Several groups around the world implemented and tested QKD systems based on variants of the BB84 protocol using either faint laser [4,5,6,7,8,9] or entangled photons [10,11,12,13,14], while, to our knowledge, only recently two groups implemented the Ekert's protocol [12,13]. In particular Naik et al.[13] implemented a variant of the Ekert's protocol based on Clauser-HorneShimony-Holt (CHSH) inequality as proposed in Ekert's paper [3], and Jennewein et al. [12] implemented the Ekert's protocol based on the Wigner inequality.In ref.[12] the Wigner inequality was first proposed to provide an easier and equally reliable eavesdropping test as the CHSH when the Ekert protocol is implemented. The necessary security proof of the Ekert protocol based on the Wigner inequality consists in verifying the violation of W ≥ 0.To obtain the Wigner inequality (W ≥ 0) it is necessary to review the Wigner argument [15]. Two main assumptions are stipulated in the proofs of the Wigner inequality: locality and realism. Locality means that Alice's measurements do not influence Bob's measurements, and vice versa. Realism means that, given any physical property, its value exists independently of its observation or measurement. The counterpart of the local-realistic theories is the non-locality behavior of quantum mechanics, a signature of quantum entanglement. In particular Wigner considered a quantum system prepared in the singlet state, and he obtained the violation of the inequality W ≥ 0, i.e. , W = −0.125. Furthermore, in the derivation of his inequality, Wigner assumed perfect anticorrelation in the measurement results. This assumption is obviously reasonable in the test of realism and locality of a physical theory (it reflects the classical counterpart of a quantum system prepared in the singlet state). Nevertheless, in terms of QKD this assumption corresponds to a lack of security.In fact, when the eavesdropper, Eve, measures photons on either one or both of Alice and Bob channels, her presence should be revealed by a higher value of W than the local-realistic theories limit W = 0, as it happens for the CHSH inequality [3]....