This paper presents a new black-box traitor tracing model where pirates use self-protection techniques to evade detection. The model assumes that pirate decoders can detect tracing attempts by analyzing the coloring of users based on traitor keys. The authors derive a necessary combinatorial condition for black-box traitor tracing in this model, showing that any system failing this condition cannot trace pirate decoders with a superlogarithmic number of traitor keys. They apply this condition to specific schemes, such as Boneh-Franklin (BF) and Kurosawa-Desmedt (KD), and show that these schemes lack black-box tracing capability in the self-protecting model unless ciphertext size is linear in the number of users. This partially resolves an open problem regarding the general black-box traceability of the BF scheme. The authors also investigate a weaker form of black-box tracing called single-query confirmation, showing that it is not possible against self-protecting pirate decoders with a superlogarithmic number of traitor keys unless the tracer has extremely high confidence. The paper concludes by analyzing concrete systems and showing that the BF scheme is incapable of black-box tracing superlogarithmic traitor collusions unless ciphertext size is linear in the number of users. The results highlight the limitations of black-box traitor tracing in the self-protecting model and provide a separation between different schemes in terms of their black-box traceability.This paper presents a new black-box traitor tracing model where pirates use self-protection techniques to evade detection. The model assumes that pirate decoders can detect tracing attempts by analyzing the coloring of users based on traitor keys. The authors derive a necessary combinatorial condition for black-box traitor tracing in this model, showing that any system failing this condition cannot trace pirate decoders with a superlogarithmic number of traitor keys. They apply this condition to specific schemes, such as Boneh-Franklin (BF) and Kurosawa-Desmedt (KD), and show that these schemes lack black-box tracing capability in the self-protecting model unless ciphertext size is linear in the number of users. This partially resolves an open problem regarding the general black-box traceability of the BF scheme. The authors also investigate a weaker form of black-box tracing called single-query confirmation, showing that it is not possible against self-protecting pirate decoders with a superlogarithmic number of traitor keys unless the tracer has extremely high confidence. The paper concludes by analyzing concrete systems and showing that the BF scheme is incapable of black-box tracing superlogarithmic traitor collusions unless ciphertext size is linear in the number of users. The results highlight the limitations of black-box traitor tracing in the self-protecting model and provide a separation between different schemes in terms of their black-box traceability.