 |
3 books on AI for Security [PDF]
September 02, 2025 | 29 |
These books explore artificial intelligence applications for detection of cyber threats, identification of vulnerabilities, surveillance, traffic control, protection of critical systems and for national security.
1. AI-Powered Software Audits: Revolutionizing Audit, Compliance, Risk, Security and Governance for Organizations
2025 by Jayant Deshmukh
Software audit - is checking the company's software for compliance with certain standards or requirements. It is usually carried out by external auditors in order to ensure licensing purity and assess security. Sometimes, fines for non-compliance are measured in millions of dollars, therefore, companies, having received a notification of an upcoming audit, usually turn on the panic mode and conduct hasty self-audits, usually manually. The author of the book tells how such companies as Google, Microsoft and JPMorgan use AI-based tools to do audits quickly and efficiently. In particular, the book describes such tools as Darktrace (uses AI to detect security breaches in real time), OneTrust (automates audits for compliance with GDPR and other regulatory requirements), SailPoint (checks identity and access management using AI), MetricStream (implements risk management using AI-based analytics)
Download PDF
2. AI, Machine Learning and Deep Learning: A Security Perspective
2023 by Fei Hu, Xiali Hei
This book focuses on the security of AI/ML models themselves, as they are now embedded in many services, systems and devices. It turns out that there are 2 types of attacks on neural networks: adversarial attacks and data poisoning attacks. In the first case, the attacker tries to change the input data in production so that the ML model misclassifies it as a different class. One of the common methods is FGSM (False Gradient Signed Norm Method) - for example, by rotating an image by an arbitrary angle. Although, such attacks can be used for various types of data: audio, text, network signals, images. Attackers try to change the data so that it still looks "normal" to a person, but the model will make mistakes. In data poisoning attacks, attackers try to manipulate the training data in order to reduce the overall accuracy of the model or cause misclassification of certain test examples or increase the training time. If the attacker targets a specific label, such attack is called targeted. It is assumed that the attacker has the ability to inject data into the training dataset. For example, this is possible when crowdsourcing data for intelligent transport systems.
Download PDF
3. AI Tools for Protecting and Preventing Sophisticated Cyber Attacks
2023 by Babulak, Eduard
This book consists of several articles on the use of AI for corporate cybersecurity. The article that interested me the most was about intrusion detection systems (IDS). It turns out that there are two main methods used for IDS: signature-based and anomaly-based. Almost all legacy intrusion detection systems are signature-based - they use rules to detect intrusions. However, for a large distributed organization, such IDS would require too many rules, which can be expensive and unreliable. If the signatures are not well defined, attackers can bypass the defenses and penetrate the system. To solve these problems, anomaly detection-based systems have been proposed - they do not require human intervention because rely on AI and trained models to improve anomaly recognition with an acceptable cost and reliability.
Download PDF
How to download PDF:
1. Install Google Books Downloader
2. Enter Book ID to the search box and press Enter
3. Click "Download Book" icon and select PDF*
* - note that for yellow books only preview pages are downloaded
1. AI-Powered Software Audits: Revolutionizing Audit, Compliance, Risk, Security and Governance for Organizations
2025 by Jayant Deshmukh
Software audit - is checking the company's software for compliance with certain standards or requirements. It is usually carried out by external auditors in order to ensure licensing purity and assess security. Sometimes, fines for non-compliance are measured in millions of dollars, therefore, companies, having received a notification of an upcoming audit, usually turn on the panic mode and conduct hasty self-audits, usually manually. The author of the book tells how such companies as Google, Microsoft and JPMorgan use AI-based tools to do audits quickly and efficiently. In particular, the book describes such tools as Darktrace (uses AI to detect security breaches in real time), OneTrust (automates audits for compliance with GDPR and other regulatory requirements), SailPoint (checks identity and access management using AI), MetricStream (implements risk management using AI-based analytics)
Download PDF
2. AI, Machine Learning and Deep Learning: A Security Perspective
2023 by Fei Hu, Xiali Hei
This book focuses on the security of AI/ML models themselves, as they are now embedded in many services, systems and devices. It turns out that there are 2 types of attacks on neural networks: adversarial attacks and data poisoning attacks. In the first case, the attacker tries to change the input data in production so that the ML model misclassifies it as a different class. One of the common methods is FGSM (False Gradient Signed Norm Method) - for example, by rotating an image by an arbitrary angle. Although, such attacks can be used for various types of data: audio, text, network signals, images. Attackers try to change the data so that it still looks "normal" to a person, but the model will make mistakes. In data poisoning attacks, attackers try to manipulate the training data in order to reduce the overall accuracy of the model or cause misclassification of certain test examples or increase the training time. If the attacker targets a specific label, such attack is called targeted. It is assumed that the attacker has the ability to inject data into the training dataset. For example, this is possible when crowdsourcing data for intelligent transport systems.
Download PDF
3. AI Tools for Protecting and Preventing Sophisticated Cyber Attacks
2023 by Babulak, Eduard
This book consists of several articles on the use of AI for corporate cybersecurity. The article that interested me the most was about intrusion detection systems (IDS). It turns out that there are two main methods used for IDS: signature-based and anomaly-based. Almost all legacy intrusion detection systems are signature-based - they use rules to detect intrusions. However, for a large distributed organization, such IDS would require too many rules, which can be expensive and unreliable. If the signatures are not well defined, attackers can bypass the defenses and penetrate the system. To solve these problems, anomaly detection-based systems have been proposed - they do not require human intervention because rely on AI and trained models to improve anomaly recognition with an acceptable cost and reliability.
Download PDF
How to download PDF:
1. Install Google Books Downloader
2. Enter Book ID to the search box and press Enter
3. Click "Download Book" icon and select PDF*
* - note that for yellow books only preview pages are downloaded
