The paper presents a device-independent security proof for Quantum Key Distribution (QKD) against collective attacks, where no assumptions are made about the specific implementation of the QKD devices or the quantum systems they operate on. The authors derive a tight bound on the Holevo information between one of the authorized parties (Alice or Bob) and the eavesdropper (Eve), as a function of the violation of a Bell-type inequality. This bound is derived using a modified version of the Ekert 1992 protocol and is shown to be independent of the quantum bit error rate (QBER). The proof is based on the non-local correlations provided by entangled particles, which cannot be reproduced by shared randomness. The authors also provide an explicit attack that saturates the bound, demonstrating its optimality. The key rate under this attack is compared to the key rate under standard QKD assumptions, showing that the device-independent approach can achieve a higher key rate. The paper concludes by discussing the implications of the proof, including its applicability to noisy or untrusted apparatuses and the detection loophole in practical scenarios.The paper presents a device-independent security proof for Quantum Key Distribution (QKD) against collective attacks, where no assumptions are made about the specific implementation of the QKD devices or the quantum systems they operate on. The authors derive a tight bound on the Holevo information between one of the authorized parties (Alice or Bob) and the eavesdropper (Eve), as a function of the violation of a Bell-type inequality. This bound is derived using a modified version of the Ekert 1992 protocol and is shown to be independent of the quantum bit error rate (QBER). The proof is based on the non-local correlations provided by entangled particles, which cannot be reproduced by shared randomness. The authors also provide an explicit attack that saturates the bound, demonstrating its optimality. The key rate under this attack is compared to the key rate under standard QKD assumptions, showing that the device-independent approach can achieve a higher key rate. The paper concludes by discussing the implications of the proof, including its applicability to noisy or untrusted apparatuses and the detection loophole in practical scenarios.