This chapter comprehensively reviews the secure, private and trustworthy learning methods from diverse perspectives. Motivated by the security and privacy issues, this chapter first presents the real data disclosure risks, attack scenarios and vulnerabilities of the current learning systems. Afterwards, the mainstream mitigation methods proposed to use differential privacy, homomorphic encryption, lightweight secure multi-party encryption, blockchain and trusted execution environment are reviewed by introducing the core algorithms and protocols. Following the tone of classical textbook writing style, in each sub-chapter, the advantages and limitations of the current schemes are discussed. Besides, this chapter argues that the adoption of single technique such as cryptograph brings about significant drawbacks. Therefore, the deep fusion of reviewed methods is promising research and practicing route to design secure while efficient learning protocols. The knowledge and insights shared in the chapter can motivate future efforts from our readers.